In this episode, we’re joined by Shubhi Mishra, founder and CEO of Raft, to talk about what it takes to become a new defense prime. Shubhi challenges the dominance of legacy primes and makes the case for smaller, faster-moving companies that can deliver what the warfighter actually needs. Through her work at Raft — a defense technology company building agile, AI-driven solutions for data fusion and rapid decision-making — she’s tackling one of the most urgent problems in defense: integrating siloed, vendor-locked systems. Shubhi shares her perspective on breaking free from “innovation theater,” reforming acquisition processes, and building real, interoperable solutions at the speed of relevance.
Read more about Shubhi here.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Unknown (00:00):
o
A'ndre Gonawela (00:09):
Hi, my name is Andre Gonawela.
Welcome back to the Burn Bagpodcast.
Today, we're going to take alittle break from all of these
geopolitical developments totalk about a key issue that
we've talked about a couple oftimes over the years, but we
haven't really delved into toodeeply, defense innovation work.
So some of you may know that Iused to work in defense
innovation.
I'm still very passionate aboutit.
(00:31):
You know, how do we get theU.S.
Armed Forces the latest andgreatest in technologies and
capabilities?
But how do we do that swiftlyHow do we do that on an
accelerated timeline?
But most importantly, how do weactually get the military what
they actually need versus thesenice-to-haves that cost billions
of dollars that then bloat ourDoD budget into this $1 trillion
(00:54):
monster?
I have a really great guesttoday.
Shubhi Mishra is joining mehere today.
She is the founder and CEO ofRaft, a cutting-edge defense
technology company deliveringAI-driven, modular solutions to
some of the U.S.
government's most complexdigital challenges.
She is quite the star herselfin DC Circle.
(01:14):
She is a two-time WASH 100awardee.
She is also a very vocaladvocate for overhauling defense
acquisition processes to breakfree from, quote, innovation
theater and empower agile,non-traditional firms She also
champions a mission-focused,buy-what-you-need approach to
modernization, emphasizingspeed, interoperability, and
(01:34):
real impact at the tacticaledge.
These are all things that Ibelieve we need to be working
more adequately towards.
So Shubhi, thanks for joiningme here today.
I really appreciate it.
Shubhi Mishra (01:45):
What a great introduction.
No, thank you so much.
Looking forward to this funconversation.
A'ndre Gonawela (01:50):
Yeah, and it's nice to also have another like
South Asian American, you know,in this podcast room.
I mean, so often, right?
Like I sort of kid that thispodcast or our guest line is
like big white wall of men whodominate our sector.
I'm South Asian myself, andit's really nice to see another,
you know, doing such incrediblework in this field.
So
Shubhi Mishra (02:09):
thank you.
No, super excited to have thisconversation.
And yeah, looking forward to avery cognitive leader of our
conversation.
A'ndre Gonawela (02:17):
Yeah.
Before we dive into some ofthese specifics, Shubhi, can you
tell us a little bit aboutyourself, a little bit about
your life, your bio, and alittle bit more?
Because I found it reallyinteresting, and I'd love to let
the audience know about that.
Shubhi Mishra (02:30):
Yeah, absolutely.
Let's see.
So I've been in this countryfor almost a decade now.
I grew up back home, which isIndia, North India, specifically
Delhi.
So I came here with a biggestand beautiful dream of, you
know, solving a lot of hardproblems and making global
(02:51):
impact and have just beenfocused on that every single
day.
And it's compounded interesthas resulted in where I am
today.
And, you know, being in the, Ifeel the best city in the world,
the impact this DC, WashingtonDC has and the entire world is
phenomenal and being surroundedby and being in the defense
(03:13):
ecosystem, you know, for all thereasons you just introduced the
audience with has beenphenomenal.
And I'm an engineer and alawyer.
I have two beautiful kids, 10and 8.
And I'm a big Peloton andtonal fanatic.
And let's say, and on vacations,I mostly, you know, go to
(03:36):
national parks and hike and goaway from the people.
And it's been fantastic.
And yeah, you know, and mypurpose in life is really all
about finding the best versionof myself and realizing and
trying to understand why I wasborn.
And in the team that I surroundmyself, I push them to do the
(03:59):
same.
So it's been an incrediblejourney thus far.
A'ndre Gonawela (04:02):
Now, that's awesome.
And when you say, you know, youtake vacations in national
parks to get away from people,when we're talking about that
bureaucracy, when we're talkingabout that acquisitions process,
boy, oh boy, are there a lot ofpeople in there.
So not to, you know, have avery odd segue, but, you know, I
think, you know, when I wasdoing some research for this
interview, when I was, you know,trying to get these questions
(04:23):
together, one of the things thatI noticed about Raft's website,
Raft being the company you run,is that when you open it, you
sort of see the words new prime.
And a lot of folks in thedefense ecosystem, they've heard
of this word prime.
But can you tell us, you know,what does prime actually mean in
the defense industry for thosewho aren't as familiar?
Shubhi Mishra (04:45):
Absolutely.
Prime to our definition ofprime is really using a lot of
people, and to solve theproblems versus products to
solve the problems.
And that is reallydifferentiated by, you know, the
scale automatically turns intohow do these prime have these
(05:16):
physical, large physicalinfrastructures from office
space to locations to theirpresence and a lot of people
with an idea and intent to solvethese mission problems.
But the gap always happens isthey are not making operators
faster, quicker, or better.
(05:38):
They are just providing morepeople to do the job, which
doesn't scale in today's era.
These
A'ndre Gonawela (05:48):
primes, for example, I don't want to name
who these primes are, obviously,in case I have them as guests
on in the future.
But when we're thinking aboutthese primes and we think about
how large they are, how manypeople there are, do you think
they're almost becoming anextension of the government in a
way, an extension of DoD withthe reliance on these types of
(06:09):
companies?
Shubhi Mishra (06:12):
I would absolutely agree.
And I think, you know, in thepast, it has worked because we
were in the peacetime, like Iwould say.
And now we're in a verydifferent time, whereas there is
a really race against a lot ofthe technology that our
adversaries have and areinvesting so much more
intentionally into to get themfar ahead.
(06:33):
And not only they are areflection of the big
bureaucracy in the governmentand, you know, the government
has bureaucracy for the rightreasons.
There are a lot of irreversibledecisions they make, and
bureaucracy helps prevent thesedecisions and makes them slow
down.
But what you need as a partneris absolutely opposite to that.
(06:53):
And what you need in a partneris scaling rapidly, fast,
pivoting until you find the bestversion of the product that you
deliver, and finding the gapthat the government cannot find.
such that you make them betterand not a mirror.
And so I think going forward,this dynamic will completely
(07:17):
change.
More new, new primes willemerge.
And with this idea of where thegap is and let's really attack
the gap versus creating amirror.
A'ndre Gonawela (07:26):
So can you tell us a little bit more about
Raft, what the company does, andespecially for those who aren't
as familiar with defensetechnology and really a lot of
these types of technologies, andhow is Raft redefining what it
actually means to be a prime?
I mean, like, what does itactually mean?
Shubhi Mishra (07:45):
So Raft is a defense technology company that
is focusing heavily inautonomous data fusion,
leveraging artificialintelligence, LLMs, limited
language models, to makeoperators super operators.
And our approach to solvingthese mission sets is completely
(08:07):
opposite to the way it's beensolved before.
We are very much focused onsolving for user needs by
sitting right next to them,rather than sitting where the
enterprise is or a PEO is andSolving for buyer needs.
(08:28):
And then in addition, rapidlyscaling to these bespoke
problems.
If you, if, so stepping backfor a second, right?
I have, Raft is verydiversified into the services it
supports from Air Force toSpace Force to Paycom, SOCOM.
And what I, and, you know, Ivisit, you these services on a
(08:53):
very frequent basis.
I'm always on the plane.
And what I've seen is everyonehas a very nuanced two problem
that they want to solve for.
And because it's very nuanced,what it results in, in creating
nuanced solutions for it.
And what the large primes arefocusing on is a one size fit
(09:15):
all solution.
And that's why they're neverable to meet the operator needs,
the end user needs.
They're able to meet buyer'sneeds because it solves for
writing, you know, one contractdocument that has all the
requirements.
But what isn't solved for islike what the guy, you know, in
the forward edge in Philippineswants.
(09:37):
It doesn't solve for that.
And the only way to solve forthat is actually going out
there, talking to the person,talking to operators and finding
a solution that is an extensionof what the product is.
And so Raft is very muchfocused on that.
And that's how we havedifferentiated ourselves from
all the others out there.
And then, you know, in the daysof, thank God for Chad GPD, you
(10:03):
know, now you don't have tofight the battle.
It's like, of course, everybodymust use AI in the day of AI,
like rapidly scaling andpivoting to find the how AI can
make a day in the life of anoperational user better has been
a game changer.
And so we are focusing so muchof that and that's how we
(10:24):
differentiate ourselves from allthe others and regularly
participating in theselarge-scale exercises that
define the area ofresponsibility for a co-com.
has been also a differentiatingfactor.
A'ndre Gonawela (10:42):
Absolutely.
And I think you mentionedsomething very important.
Companies need to be sittingwith the actual warfighters, the
folks who are actually at thatforward edge to identify the
needs, the requirements, thepain points, really.
And I mean, When you'rethinking about the DoD
acquisitions process, there areso many layers, so many senior
(11:04):
officials, commanders, officerswho you need to get the approval
of.
You need to satisfy the variousstakeholders.
You need to get this actionofficers involved.
You need to get the seniorofficers approval and so on.
And that's such a battle to getthrough.
So I mean, what does itactually take to build something
the warfighter actually wantsand needs versus filling a
(11:29):
request by a military entitythat wants something with AI
just to have something with AIbecause it's a sexy new
technology, right?
Because sometimes I feel like,especially in my past
experience, we'll have theseentities who want these new
technologies, but there won'treally be a problem to solve
with it, they'll sort of moldthe problem around the solution
(11:53):
and sort of force fit it inversus the other way around,
finding a solution that fits theproblem.
Shubhi Mishra (11:59):
In my opinion, DoD and government at large,
it's a very consensus-baseddecision-making process.
And there's no, yes, there's asole decision maker with whoever
the authority lies, but thesole decision maker depends on a
lot of people to get andunderstand that decision and
(12:20):
make sure it's the rightdecision.
And so I think the answer isand and not or.
It's what we have done and beensuccessful at.
And this is something I'm oftenasked of new upcoming companies
when they seek advice is, And Ithink you cannot alienate the
buyer community or share withthem how wrong sometimes they
(12:44):
could be.
And while making sure that youare championing for the users
because Over a period of time,it catches up.
Either your product is beingused or it's not being used.
Either it's being talked aboutin the circles or it's not being
talked about.
And the question just becomes,you know, this large framework
(13:04):
of abstract of meet theserequirements.
How do you really tailor themand be creative such that you
can champion the user needsthrough that?
So it ultimately comes down tothat.
But I think it's an endquestion and end answer, I
should rather say, rather thanHow do you choose to do one or
the other?
(13:24):
And the end answer is, youknow, learned it the hard way is
it's about, it should,unfortunately, and that's what
for the vendors, a lot of thecycles are spent on that they
rather not be spent on, but itis part of playing the game is
you got to make sure it meetseverybody's checklist to the
maximum level.
So
A'ndre Gonawela (13:44):
you've spoken out against innovation theater
in defense.
What does that term actuallymean to you?
And how do we get beyondinnovation theater?
Shubhi Mishra (13:52):
Yeah, so it's the performance.
You and I were talking about ita little bit before, but it's
the performance to make everyonearound you seem or believe that
action is resulting in impact.
I think innovation theater issometimes needed to get the
(14:14):
marketing cycles in, but then Ithink the shift needs to happen
right after or sequentially orparallelly about how do you make
this, what you're talking aboutinto something a user can use
on the farthest edge and have itdelivered operational value.
(14:36):
And it's easy to say, andeverybody has great intent, But
it's so hard to do.
So I'll give you an example.
I just came back fromIndo-Pekong, Hawaii, and I was
on the island for a couple ofdays, you know, just meeting
with the decision makers.
And I some of the time is spentat the AOC and our folks that
(14:59):
are working out there, workingwith, you know, people in
uniform.
So you have this old, you know,World War era buildings that
you can imagine the state,right?
There's barely coffee, even avending machine.
There's nothing like thatexists.
And then to you, that's yourskiff.
That's where you live out of.
And then you come out in theblazing Hawaii sun.
Speaker 01 (15:22):
Oh,
A'ndre Gonawela (15:23):
gee, yeah.
Shubhi Mishra (15:24):
you know, without any fans, but it's a ticky bar,
which is kind of cool, youknow, with like those
old-fashioned shades on them.
You come out, you know, youcatch up on life or you use the
unclassified version of acomputer to check on any updates
or how to fix a problem.
(15:44):
And then you go back in and dothe same thing.
And you do that day in and dayout.
And that is not easy.
But that's necessary to turnthat in into operational value.
And I think the last part, thedoing the do and the hard do is
where most of us fall shortbecause it's humanly, it's
(16:05):
humanly hard, physically hard.
A'ndre Gonawela (16:06):
No, yeah, exactly.
And I mean, like when we'retalking about, you know, what do
we actually do?
work on produce to gain thatoperational value?
I mean, what types of metricsor outcomes are we using to
determine if that capability istruly impactful versus being
more of that performativeinnovation?
I mean, I've spent a couple ofyears working in the defense
(16:27):
innovation space, and nowadays Isort of cringe when I hear the
word innovation being thrownaround by so many different
people who'll sort of use thatterm to just latch it onto
anything new.
So
Shubhi Mishra (16:40):
I think it starts with participating in
exercises, these exercises whichare replicating how a battle or
war may be fought.
And it starts with that, and itstarts with, can you do with
less humans?
Can you do with not awhiteboard?
How much can you automateend-to-end?
(17:02):
And the result of that ends upwith more and more operational
users, operators using yourtool, And then the word of mouth
circulating across the board.
So now that we have, let's say,a particular AOC tomorrow, we
have another one added.
And added because they want toand they want to use and they
want to make their life better,not because it's coming top
(17:24):
down.
And I think that is the metricthat we keep a close eye on.
And I highly recommend, youknow, businesses to do that.
And that requires a lot ofinvestment.
So instead of, you know, thisis, again, one of the
conversations I'm asked often,instead of investing in maybe a
large growth team or, you know,business development team,
(17:44):
invest in finding ways toparticipate in exercises and
really demonstrating valuebecause there's a wide gap right
now that exists.
And if you can make anybody'slife easier, they will not let
you go.
A'ndre Gonawela (18:00):
Oh, absolutely.
So I want to move theconversation into a little bit
about the work you do at Raft,because it's fascinating work.
I'm still trying to wrap myhead around it as I'm not a
technologist, so I'll definitelyappreciate your help in trying
to understand this.
But you talk about data fusion,autonomous data fusion, how
(18:21):
it's a big critical challenge intoday's defense environment.
Can you tell us a little bitmore about what autonomous data
fusion is?
Shubhi Mishra (18:30):
now data fusion is being done by human beings
using their gray matter and ifwe want to get to the level we
want to and compete with ouradversaries who by the way have
not only figured out end-to-endsystems and and can where
everything talks to everythingso it's just faster because of
(18:50):
the nature of it but also arecontinually investing in
technology um we must shift fromhumans gray matter to doing the
do to machines doing most of itautonomously and such that a
human is observing it and makingsure it's doing the right way.
A'ndre Gonawela (19:11):
So this is like the conversion of different
types of data into one universallanguage.
Shubhi Mishra (19:17):
So it extends a little bit beyond that, right?
So that is the data fusionprocess.
but autonomously requiresmachine learning models in an AI
to not only read what differentformats, standards, schema,
(19:37):
whatever you call it, whateveris coming in, and then translate
it back to how an operatorwould want it in how they
consume the information.
A, you know, a, somebody, a paycomp, somebody's area of
responsibility within a packhalf may be different from a
(19:57):
pack fleet.
And they, they consumeinformation different way, but
they need similar information.
These systems that they'reusing or consuming information
from are, have been built indifferent decades by different
vendors and, and, you know,which are not necessarily
talking to each other.
So it's always, how do you makesense of all this stuff into
(20:20):
something that the operator hasbeen trained on?
And the way different servicesgrow and teach their operators
very differently, and you haveto autonomously make sure the
information they're consumingmakes sense to them.
A'ndre Gonawela (20:37):
So it's like you have all of these different
capabilities, these differentplatforms, these different types
of technologies, computers,whatever, who are all outputting
all of these different types ofdata.
Say you have a team of people,one speaking Hindi, one speaking
Spanish, one speaking English,one speaking German, and you
(20:58):
have all of this great data, butthe one person who needs to
make a decision on it cannoteasily understand it, cannot
easily translate that into oneuniversal language like English
or something.
And we need to do thisautomatically.
So it's that sort of a veryrough allegory.
Shubhi Mishra (21:17):
Yeah, that's exactly it.
And what you said at the end issuper key.
It's like it is using theirlanguage, whether that is
English or something else.
And using natural language.
So it's not necessarily as ahuman as you would interact with
something.
And that's where the autonomouspart comes in, the natural
(21:38):
language part comes in.
So far, the tools that existand have been deployed are...
not natural language tools thatare not a human talking to
another human tool.
So imagine now, go back,imagine when operators are doing
their job.
When they pick up the phone andcall the other guy sitting near
the skiff, they're using humanlanguage.
(21:59):
They don't use machinelanguage.
But the tools right now thatexist, it's a machine language
talking to a human.
So if you want to change thebehavior and if you want to
change AI being the forcingfunction that enables them, we
gotta replicate this phone callto the upper operator and make
it frictionless and human talk.
(22:20):
So that's where the autonomouspart comes in and the data
fusion is just around themachine to machine talk.
A'ndre Gonawela (22:27):
Yeah, and I mean, when I'm thinking about
why that's so important, I'mthinking about all of the
different things that the DoD isprocuring, whether it's fighter
jets, whether it's sensors,whether it's the new unmanned
capabilities, whether it's allof these different types of
technologies, folks, that we'rethinking about that one may use
(22:47):
in a singular operation, youknow, when we're bringing one of
those capabilities in, we wantto integrate that into the
broader force to make sure itall sort of fits seamlessly.
And, you know, I'm noticingthis, you know, not just in a US
context, but like a lot of ourAsian partners, for example,
right?
They're like procuringdifferent types of capabilities,
(23:08):
right?
Like, you know, India is buyingfighter jets from France for
example, the Rafales.
Are those Rafale fighter jetsfitting into the broader
operational capabilities of theIndian Army?
Are they sort of making surethat all the data can sort of
talk to each other?
And, you know, I'm definitelyobserving that as we're
(23:28):
expanding the aperture ofacquisitions beyond the U.S.,
but into the allies, with theallies, especially with things
like AUKUS, with things like,you know, U.S.-India defense
cooperation and so on.
Is that...
Correct.
Shubhi Mishra (23:41):
You're 100% on point there.
And it's almost like you'resolving for the symptom.
symptom of new technology, newtechnology, but we are not
talking about the root cause.
And the root cause is, thereare only two ways to solve it.
One, the way China solved it.
It's one thing.
It's one company, onesubsidiaries of that company,
(24:03):
and they must talk to eachother.
And that's what I call aboutsolving end-to-end.
And then in America, wherethere's a multi-vendor ecosystem
and the best may win, we don'tfocus on solving this root
cause.
Yes, best may win.
Right now, humans are enablingthis talk track between all
these different systems.
What happens in the future whenseconds result in life and
(24:27):
death?
And that's not a conversationyet because we're still warming
up to it.
And I think that's the mostimportant conversation we need
to have.
A'ndre Gonawela (24:35):
No, yeah, absolutely.
Because I mean, you know, inthe current defense innovation
ecosystem, we still have a lotof work to do, but we have so
many interesting companies, somany interesting technologies,
so many new things that arepopulating, you know, our
capabilities.
But how are we connecting thatefficiently?
Because that's a key issuebecause, you know.
Shubhi Mishra (24:52):
It's the boring part, but the most important
part.
A'ndre Gonawela (24:55):
But it's vitally important, especially
when you're in the battlefield,when you have, you know, only
seconds to make a decision oflife and death, you need
processing power, right?
to automatically do this.
Because I mean, again, like,you know, folks, like going back
to the allegory about you havethe team of different people who
speak different languages.
If you have a human doing it,it's going to take so long to
actually do that translating byhand.
(25:16):
I mean, folks go back to yourFrench classes, right?
And think about how hard thatwas.
I mean, if you had a computerdoing it, it'd be fantastic.
So, I mean, you know, whenwe're talking about, you know,
universal data fusion, what doesthat actually mean?
look like in practice and whyis it so hard
Shubhi Mishra (25:36):
yeah i mean see it's hard because technically
it's hard right i mean to solvefor it's just decades of
different vendors decades ofdifferent systems and and the
way dod has wanted to solve forit to make a standard
standardized things and theproblem with that is and that's
(25:57):
why there's not a lot of buy-inand people who are making these
decisions don't understand theground truth, which is Space
Delta, Space Force Delta issolving for their specific area
of responsibility.
They will create something thathelps them get that information
(26:21):
faster and quicker.
On the other hand, you know,some organization within Air
Force or Navy, they will solvefor very specific ways they want
to solve for.
And so could they adopt thisstandard?
Yeah.
Would that result in themlosing precious time to enable
(26:41):
their decisions?
Absolutely.
And so this one standard, sizefit all, has not been adopted.
And I think it should be thereverse of it.
You do whatever you need to,but let's make sure the data
fusion autonomously happens.
So I think technically that'sone big challenge, just really
hard engineering.
The second is it isbureaucratically challenged and
(27:08):
there are a lot of policies thatstand in the way.
And those policies weredesigned for good reason.
And however, there are too manypolicies now and nobody wants
to share data with the otherentity because, I don't know.
To me, those reasons don't makeany sense, but they are
(27:32):
politically and bureaucraticallychallenges.
And I think a lot of those arealso fed by their different
vendors wanting to protect theterritory and making sure they
can protect the territory ifthey don't share data.
That's a huge problem.
piece of it.
And I also think the thirdpiece is, I don't think Congress
(27:53):
has looked at this problemstatement super closely.
I've spent a lot of time on theHill, continuously do so.
And I'm amazed at the gap theyhave between what the
information they know and gatherto what the information is on
the field.
And that's where, you know,they've appreciated my insight
(28:14):
because I'm able to translateand get, they can hear it from
the horse's mouth or almost fromthe horse's mouth of the gap.
And I think they haven't askedthese questions.
They haven't focused on this somuch and they just get pitched
how new things will solve thisproblem forever versus we must
(28:36):
find a way for the old and newto come together.
So more focus from Congresswould also help this.
But that's what I would say,the three challenges.
Technically bureaucratic andnot enough oversight,
congressional oversight.
A'ndre Gonawela (28:53):
Oh, yeah.
No, no, for sure.
And I mean, like, how do you...
I mean, when we're talkingabout how you actually do this,
like, how do you actuallyapproach integrating these
disparate data systems, youknow, from these multiple
vendors?
Like, what's the key to this?
Shubhi Mishra (29:07):
So...
Technically, there are thingsthat, at least the way we have
approached it, is reallydeveloping an abstract layer.
And so, you know, that givesyou 70, 80% of the way.
And then some of these thingsare...
Just similar type formats andso text format, you know, video
(29:32):
or pictures and just puttingthem in bigger buckets and then
figuring out how to parsethrough that.
So that's one way technicallywe have done it.
And the next is really a lot ofpartnership with other industry
partners.
And what that helps us is, andI would say any vendor helps is,
it fastens the getting to theend phase.
(29:53):
quicker and we get there fasterjust because then you don't
have the additional bureaucracyof just DoD and government
coming in the middle and you canjust make B2B connections.
And so their system can talk toyour system and, you know, you
get to the end result faster.
(30:13):
So that's how I would say wehave approached it.
A'ndre Gonawela (30:15):
No, for sure.
Definitely.
And, you know, we talked aboutthe information gaps belying our
congressional representatives.
We talked about, you know, whythere are gaps between the
forces and sort of thestylification of that data.
But do you think theacquisition community itself, I
mean, fully understands theimplications of fragmented data
(30:37):
architectures?
Shubhi Mishra (30:39):
I wish they do.
I wish they would.
They're not living the pain.
A'ndre Gonawela (30:44):
Why not?
Shubhi Mishra (30:45):
The pain must be felt.
And there's only one reason tofeel this pain is go out to
Guam, go out to Philippines andsee and experience a day in life
of that operator.
I think that will be gamechangers.
And I feel the ones acquisitionprofessionals that I've come
across who have in uniform, Ispecifically say that have
(31:09):
rotated out.
or have experience towards thefar edges are the most, provide
exponential results for thesecommunities because they
understand.
Yeah,
A'ndre Gonawela (31:22):
absolutely.
So I mean, When we're thinkingabout all of these challenges,
not just within autonomous datafusion, but sort of some of the
larger challenges, I mean, youallude to, I think, what you
just said, right, about theacquisition community, those
officials not reallyunderstanding the pain because
(31:42):
they don't feel the actual pain.
How do we work towards this?
I mean, what does the nextgeneration defense contractor
do?
look like?
How do they navigate thisintense, complicated, soulless
acquisitions framework with allof these different entities,
(32:04):
these different people you needto communicate the pain to, make
sure you're not communicatedwithin a silo and so on?
How do contractors navigatethis?
Shubhi Mishra (32:14):
So I think there is a now what we need to do, and
I think there'll be a differentfuture in a few years from now.
The now is unfortunately, wegot to bring the two communities
together and really have theusers advocate for what they
want.
And that's the other big gapI've seen is users don't know or
(32:38):
have a blueprint of how to evenask for what they want.
And that's where vendors likeus can help them because we know
how the game is being played.
I say game.
I have a lot of respect forthis game.
It's a beautiful game.
It's
A'ndre Gonawela (32:57):
a game nonetheless.
Shubhi Mishra (32:58):
Yes, yes.
But I do think it is educatingand enabling these operators on
the edge to go to theacquisition community to get
them what they need.
And it's continuously a lot ofconversations.
I also think the...
(33:19):
It is the only model we haveseen for success for any vendors
in this ecosystems are the bigprime or the legacy model.
And I think the new primesshould absolutely, and it's not
intuitive, it's against thehuman intuition, should
(33:41):
absolutely fight against thatand make sure that they don't
end up being bloated.
Because bureaucracy burns cash.
A'ndre Gonawela (34:20):
overly bureaucratic organization?
Shubhi Mishra (34:23):
Three letters.
ABC, always be cutting.
Cut what doesn't work rapidly,invest in what's working.
I think that's the simplest wayto structure it.
And the beauty of now withthese AI tools is you don't need
that much manpower, for lack ofa better word, to do a lot of
(34:47):
these things.
I think...
Invest in the people that wantto multiply and are 10x and not
invest in things and processesthat don't work.
A'ndre Gonawela (34:59):
Yeah.
And I mean, you know, I sort ofnoticed this with like a lot of
these companies that have beenstartups that sort of start off
small as they have success, asthey grow.
And you're in this sort of thisweird space where you're a
midsize company.
Growth is accelerating.
And then as a result, youalmost feel the need to put in
(35:20):
bureaucracy, right, to make surethat you can manage your
organization well.
I mean, people will often saythat being small and agile.
is the advantage.
However, how do you scale up anorganization to fill these
needs, to grow the success andso on?
Where's the middle ground interms of that management?
Shubhi Mishra (35:43):
So I'll tell you.
So it's one of those thingswhere, look, if a company is
selling that their capabilityand products is all around
enabling operators throughartificial intelligence, machine
learning, and data, they betterbe doing that on their side of
(36:04):
things, which means is they mustover-invest in these tools for
themselves and make 1x to 10x.
And so the same thing appliesbecause that's the premise of
everything.
And in my opinion, it's alwaysgoing to be, unintuitive because
(36:29):
it's so new, the capabilitiesthese AI is providing you.
But I think fight that urge tohave a bunch of middle managers.
Fight the urge for you need ahuman to help a human to help a
(36:49):
human.
Fight that.
That just slows everythingdown.
You have too many decisionmakers.
And I approach it verydifferently where it's who has
the veto versus the decision,right?
More confusion results stopsmoving fast and stops agileness.
(37:13):
And I understand this idea ofmiddle-sized company versus
small company versus largecompany.
It's measured by revenue, but Ithink a lot of that is
changing.
I do think in today's world,it's all about what value you're
creating and how you're solvingthe mission.
Thanks to not only the recentadvances in AI, but also the new
(37:39):
government in charge, they'revery much focusing on what
capability do you provide versushow large your team size is.
And given all of that, I reallythink it should be very much
focused on Solving the missionneeds, that creates a lot of
(37:59):
value versus, I don't know,scale with millions of dollars
that are humans that will bereplaced any day by AI in any
case.
A'ndre Gonawela (38:12):
Yeah.
And I mean, when we're talkingabout AI, I mean, you know, a
lot of people have a vision forhow AI can integrate into our
capabilities, but it's a veryvague vision blocked by a lot of
fog and a lack of clarity.
But, you know, there's also alot of caution in how AI
(38:32):
integrates into our defensecapabilities.
I mean, it should be...
does good AI adoption anddefense actually look like?
And what should it not looklike?
Shubhi Mishra (38:42):
So the way I think about it is just AI in
terms of decision-making lens.
As a human being, there are twotypes of decisions you end up
making.
As reversible, doesn't causelife and death.
Irreversible, your action ordecision will result in life or
death.
Majority of the decisions thatare being made within the
(39:03):
ecosystem are reversible.
You can change your decision ifit's not the right decision.
You can go back.
And I think that is absolutelyprime for anything artificial
intelligence brings in.
Where your decision causes lifeand death, we must proceed with
a lot of caution.
There's no coming back fromthere.
(39:23):
There we...
must test this.
There are a lot of exercises.
There are a lot of testing thatneeds to happen.
Even after that, there alwaysneeds to be an operator on the
loop or the human on the loop tomake sure the decision is the
right decision.
And furthermore, if you go intoit more, it's just not about AI
(39:45):
tools.
It's about AI tools that areshowing you how they're making
the decisions.
There are a lot of tools outthere right now in the
ecosystem.
You give them a question,they'll give you an answer.
That's not the right way toapproach it.
For a human and machine tobuild trust, the machine must
show how it's thinking and howit got to the answer.
(40:07):
And over time, you will buildtrust with it.
You like it or not like it, oryou tweak it.
But at the end of the day,showing that thinking is going
to be critical for thisintegration of decisions that
can be reversed.
A'ndre Gonawela (40:25):
Absolutely.
And I mean, how do we alsoensure the ethical,
operationally relevant andsecure AI integration in
national security missions?
Because I mean, that's some ofthe biggest concerns that people
will have about AI.
I,
Shubhi Mishra (40:40):
again, think of it as the two buckets of, you
know, if your decision can bereversed, then Let's weigh it
less in terms of these stagegates we have of something being
ethical or non-ethical.
Of course, nothing should benon-ethical, and that's why we
should be testing this out.
(41:00):
But over-emphasis on it,massive regulations on it when
it is an irreversible decisionand a life-or-death decision.
A'ndre Gonawela (41:09):
Gotcha.
So, I mean, you know, as wesort of run this interview out,
I mean, you've said fear is thegreatest enemy of progress.
I mean, how does that informyour own leadership style at
Raft?
Shubhi Mishra (41:20):
So I've always said run towards fear with a lot
of courage.
And I think as human beings,we, our monkey brain, our
amygdala brain, it's a fight andflight and the minute something
(41:41):
hard comes in, you just want toflee.
And I think that's the signalto run towards the hard problem,
run towards your fear with alot of courage.
And that is something we doevery single day.
And that is the sole reason whyraft is where it is and where
it stands.
(42:01):
And I, that is the sole reasonfor the curve of past six years
for us has been so exponentialthat we are sought out by our
customers to come help themversus the other way around.
And if you're not runningtowards fear with a lot of
(42:21):
courage, you're not having fun,right?
A'ndre Gonawela (42:23):
I mean, how does one have fun amidst all the
friction and the inertia ofthis defense bureaucracy?
Because I mean...
It sounds like a whole lot ofnot fun.
Shubhi Mishra (42:36):
It is.
Oh, the mundane parts areboring.
And yes, those are just thingsthat need to be done.
But it's so much fun to forge anew path and to redefine the
word prime, to really help theseoperators who need help.
There's so much...
(42:57):
Meaning, I mean, I'm going togo a little philosophical.
There's so much meaning in lifefor that.
I mean, you know, when youthink about all the time you
spent and all the nights youstayed up, and if you realize
that this is the impact one hadon the journey of a country that
(43:21):
impacts the world at thatscale, it's beautiful.
Couldn't be more fun.
A'ndre Gonawela (43:29):
So what changes, I think, technological
or institutional, are you mostexcited about over the next 12
to 18 months?
Because I mean, you know,within the span of like one
year, we see so manytechnological advances.
I mean, if we're looking atChina, we saw deep seek
suddenly.
come onto the world stage andthat set off a firestorm of
shock and awe and so many otherthings.
(43:51):
And, you know, we're in a newadministration now.
There are a lot ofinstitutional changes occurring.
What's exciting you the mostabout these types of changes
we're seeing in the short term?
Shubhi Mishra (44:03):
I will say, yes, there's a lot of conversation in
the media, in theadministration, on the hill
around AI, but it is not gettingto the edge yet.
It is not getting to the peoplethat can use it to make their
lives better.
And so I'm super excited to seethe adoption and the trust that
(44:29):
the operators start buildingwith these AI systems.
That's one.
And the second thing is there'sa lot of PowerPoint talk still,
a lot of innovation theater.
So, you know, that's the otherpart.
Now that the innovation theatercan act fast and can result in
action more because the time isnow, I'm excited to see those
(44:54):
PowerPoints being turned intovaluable tools that can change
lives.
A'ndre Gonawela (44:59):
No, absolutely.
Shubhi, thank you for joiningme here today.
This was a really greatconversation.
I really appreciate it.
I really appreciatedconnecting.
on, I think, artist Dane forinnovation theater, because I
think that's a huge blocker ofprogress.
But I mean, you know, I reallyfound the conversation on Data
Fusion really fascinating.
And for our audience members,definitely check out Raft's
(45:22):
work.
Data fusion is a very importanttopic if you're interested in
defense force modernization.
But also, I mean, think aboutShubhi's lessons on leadership
style and organization and soon.
But Shubhi, thanks for joiningme here today.
I really appreciate it.
t
Shubhi Mishra (45:40):
It was fantastic, Andre.
Thanks so much.
Thank you.
o
A'ndre Gonawela (00:09):
Hi, my name is Andre Gonawela.
Welcome back to the Burn Bagpodcast.
Today, we're going to take alittle break from all of these
geopolitical developments totalk about a key issue that
we've talked about a couple oftimes over the years, but we
haven't really delved into toodeeply, defense innovation work.
So some of you may know that Iused to work in defense
innovation.
I'm still very passionate aboutit.
(00:31):
You know, how do we get theU.S.
Armed Forces the latest andgreatest in technologies and
capabilities?
But how do we do that swiftlyHow do we do that on an
accelerated timeline?
But most importantly, how do weactually get the military what
they actually need versus thesenice-to-haves that cost billions
of dollars that then bloat ourDoD budget into this $1 trillion
(00:54):
monster?
I have a really great guesttoday.
Shubhi Mishra is joining mehere today.
She is the founder and CEO ofRaft, a cutting-edge defense
technology company deliveringAI-driven, modular solutions to
some of the U.S.
government's most complexdigital challenges.
She is quite the star herselfin DC Circle.
(01:14):
She is a two-time WASH 100awardee.
She is also a very vocaladvocate for overhauling defense
acquisition processes to breakfree from, quote, innovation
theater and empower agile,non-traditional firms She also
champions a mission-focused,buy-what-you-need approach to
modernization, emphasizingspeed, interoperability, and
(01:34):
real impact at the tacticaledge.
These are all things that Ibelieve we need to be working
more adequately towards.
So Shubhi, thanks for joiningme here today.
I really appreciate it.
Shubhi Mishra (01:45):
What a great introduction.
No, thank you so much.
Looking forward to this funconversation.
A'ndre Gonawela (01:50):
Yeah, and it's nice to also have another like
South Asian American, you know,in this podcast room.
I mean, so often, right?
Like I sort of kid that thispodcast or our guest line is
like big white wall of men whodominate our sector.
I'm South Asian myself, andit's really nice to see another,
you know, doing such incrediblework in this field.
So
Shubhi Mishra (02:09):
thank you.
No, super excited to have thisconversation.
And yeah, looking forward to avery cognitive leader of our
conversation.
A'ndre Gonawela (02:17):
Yeah.
Before we dive into some ofthese specifics, Shubhi, can you
tell us a little bit aboutyourself, a little bit about
your life, your bio, and alittle bit more?
Because I found it reallyinteresting, and I'd love to let
the audience know about that.
Shubhi Mishra (02:30):
Yeah, absolutely.
Let's see.
So I've been in this countryfor almost a decade now.
I grew up back home, which isIndia, North India, specifically
Delhi.
So I came here with a biggestand beautiful dream of, you
know, solving a lot of hardproblems and making global
(02:51):
impact and have just beenfocused on that every single
day.
And it's compounded interesthas resulted in where I am
today.
And, you know, being in the, Ifeel the best city in the world,
the impact this DC, WashingtonDC has and the entire world is
phenomenal and being surroundedby and being in the defense
(03:13):
ecosystem, you know, for all thereasons you just introduced the
audience with has beenphenomenal.
And I'm an engineer and alawyer.
I have two beautiful kids, 10and 8.
And I'm a big Peloton andtonal fanatic.
And let's say, and on vacations,I mostly, you know, go to
(03:36):
national parks and hike and goaway from the people.
And it's been fantastic.
And yeah, you know, and mypurpose in life is really all
about finding the best versionof myself and realizing and
trying to understand why I wasborn.
And in the team that I surroundmyself, I push them to do the
(03:59):
same.
So it's been an incrediblejourney thus far.
A'ndre Gonawela (04:02):
Now, that's awesome.
And when you say, you know, youtake vacations in national
parks to get away from people,when we're talking about that
bureaucracy, when we're talkingabout that acquisitions process,
boy, oh boy, are there a lot ofpeople in there.
So not to, you know, have avery odd segue, but, you know, I
think, you know, when I wasdoing some research for this
interview, when I was, you know,trying to get these questions
(04:23):
together, one of the things thatI noticed about Raft's website,
Raft being the company you run,is that when you open it, you
sort of see the words new prime.
And a lot of folks in thedefense ecosystem, they've heard
of this word prime.
But can you tell us, you know,what does prime actually mean in
the defense industry for thosewho aren't as familiar?
Shubhi Mishra (04:45):
Absolutely.
Prime to our definition ofprime is really using a lot of
people, and to solve theproblems versus products to
solve the problems.
And that is reallydifferentiated by, you know, the
scale automatically turns intohow do these prime have these
(05:16):
physical, large physicalinfrastructures from office
space to locations to theirpresence and a lot of people
with an idea and intent to solvethese mission problems.
But the gap always happens isthey are not making operators
faster, quicker, or better.
(05:38):
They are just providing morepeople to do the job, which
doesn't scale in today's era.
These
A'ndre Gonawela (05:48):
primes, for example, I don't want to name
who these primes are, obviously,in case I have them as guests
on in the future.
But when we're thinking aboutthese primes and we think about
how large they are, how manypeople there are, do you think
they're almost becoming anextension of the government in a
way, an extension of DoD withthe reliance on these types of
(06:09):
companies?
Shubhi Mishra (06:12):
I would absolutely agree.
And I think, you know, in thepast, it has worked because we
were in the peacetime, like Iwould say.
And now we're in a verydifferent time, whereas there is
a really race against a lot ofthe technology that our
adversaries have and areinvesting so much more
intentionally into to get themfar ahead.
(06:33):
And not only they are areflection of the big
bureaucracy in the governmentand, you know, the government
has bureaucracy for the rightreasons.
There are a lot of irreversibledecisions they make, and
bureaucracy helps prevent thesedecisions and makes them slow
down.
But what you need as a partneris absolutely opposite to that.
(06:53):
And what you need in a partneris scaling rapidly, fast,
pivoting until you find the bestversion of the product that you
deliver, and finding the gapthat the government cannot find.
such that you make them betterand not a mirror.
And so I think going forward,this dynamic will completely
(07:17):
change.
More new, new primes willemerge.
And with this idea of where thegap is and let's really attack
the gap versus creating amirror.
A'ndre Gonawela (07:26):
So can you tell us a little bit more about
Raft, what the company does, andespecially for those who aren't
as familiar with defensetechnology and really a lot of
these types of technologies, andhow is Raft redefining what it
actually means to be a prime?
I mean, like, what does itactually mean?
Shubhi Mishra (07:45):
So Raft is a defense technology company that
is focusing heavily inautonomous data fusion,
leveraging artificialintelligence, LLMs, limited
language models, to makeoperators super operators.
And our approach to solvingthese mission sets is completely
(08:07):
opposite to the way it's beensolved before.
We are very much focused onsolving for user needs by
sitting right next to them,rather than sitting where the
enterprise is or a PEO is andSolving for buyer needs.
(08:28):
And then in addition, rapidlyscaling to these bespoke
problems.
If you, if, so stepping backfor a second, right?
I have, Raft is verydiversified into the services it
supports from Air Force toSpace Force to Paycom, SOCOM.
And what I, and, you know, Ivisit, you these services on a
(08:53):
very frequent basis.
I'm always on the plane.
And what I've seen is everyonehas a very nuanced two problem
that they want to solve for.
And because it's very nuanced,what it results in, in creating
nuanced solutions for it.
And what the large primes arefocusing on is a one size fit
(09:15):
all solution.
And that's why they're neverable to meet the operator needs,
the end user needs.
They're able to meet buyer'sneeds because it solves for
writing, you know, one contractdocument that has all the
requirements.
But what isn't solved for islike what the guy, you know, in
the forward edge in Philippineswants.
(09:37):
It doesn't solve for that.
And the only way to solve forthat is actually going out
there, talking to the person,talking to operators and finding
a solution that is an extensionof what the product is.
And so Raft is very muchfocused on that.
And that's how we havedifferentiated ourselves from
all the others out there.
And then, you know, in the daysof, thank God for Chad GPD, you
(10:03):
know, now you don't have tofight the battle.
It's like, of course, everybodymust use AI in the day of AI,
like rapidly scaling andpivoting to find the how AI can
make a day in the life of anoperational user better has been
a game changer.
And so we are focusing so muchof that and that's how we
(10:24):
differentiate ourselves from allthe others and regularly
participating in theselarge-scale exercises that
define the area ofresponsibility for a co-com.
has been also a differentiatingfactor.
A'ndre Gonawela (10:42):
Absolutely.
And I think you mentionedsomething very important.
Companies need to be sittingwith the actual warfighters, the
folks who are actually at thatforward edge to identify the
needs, the requirements, thepain points, really.
And I mean, When you'rethinking about the DoD
acquisitions process, there areso many layers, so many senior
(11:04):
officials, commanders, officerswho you need to get the approval
of.
You need to satisfy the variousstakeholders.
You need to get this actionofficers involved.
You need to get the seniorofficers approval and so on.
And that's such a battle to getthrough.
So I mean, what does itactually take to build something
the warfighter actually wantsand needs versus filling a
(11:29):
request by a military entitythat wants something with AI
just to have something with AIbecause it's a sexy new
technology, right?
Because sometimes I feel like,especially in my past
experience, we'll have theseentities who want these new
technologies, but there won'treally be a problem to solve
with it, they'll sort of moldthe problem around the solution
(11:53):
and sort of force fit it inversus the other way around,
finding a solution that fits theproblem.
Shubhi Mishra (11:59):
In my opinion, DoD and government at large,
it's a very consensus-baseddecision-making process.
And there's no, yes, there's asole decision maker with whoever
the authority lies, but thesole decision maker depends on a
lot of people to get andunderstand that decision and
(12:20):
make sure it's the rightdecision.
And so I think the answer isand and not or.
It's what we have done and beensuccessful at.
And this is something I'm oftenasked of new upcoming companies
when they seek advice is, And Ithink you cannot alienate the
buyer community or share withthem how wrong sometimes they
(12:44):
could be.
And while making sure that youare championing for the users
because Over a period of time,it catches up.
Either your product is beingused or it's not being used.
Either it's being talked aboutin the circles or it's not being
talked about.
And the question just becomes,you know, this large framework
(13:04):
of abstract of meet theserequirements.
How do you really tailor themand be creative such that you
can champion the user needsthrough that?
So it ultimately comes down tothat.
But I think it's an endquestion and end answer, I
should rather say, rather thanHow do you choose to do one or
the other?
(13:24):
And the end answer is, youknow, learned it the hard way is
it's about, it should,unfortunately, and that's what
for the vendors, a lot of thecycles are spent on that they
rather not be spent on, but itis part of playing the game is
you got to make sure it meetseverybody's checklist to the
maximum level.
So
A'ndre Gonawela (13:44):
you've spoken out against innovation theater
in defense.
What does that term actuallymean to you?
And how do we get beyondinnovation theater?
Shubhi Mishra (13:52):
Yeah, so it's the performance.
You and I were talking about ita little bit before, but it's
the performance to make everyonearound you seem or believe that
action is resulting in impact.
I think innovation theater issometimes needed to get the
(14:14):
marketing cycles in, but then Ithink the shift needs to happen
right after or sequentially orparallelly about how do you make
this, what you're talking aboutinto something a user can use
on the farthest edge and have itdelivered operational value.
(14:36):
And it's easy to say, andeverybody has great intent, But
it's so hard to do.
So I'll give you an example.
I just came back fromIndo-Pekong, Hawaii, and I was
on the island for a couple ofdays, you know, just meeting
with the decision makers.
And I some of the time is spentat the AOC and our folks that
(14:59):
are working out there, workingwith, you know, people in
uniform.
So you have this old, you know,World War era buildings that
you can imagine the state,right?
There's barely coffee, even avending machine.
There's nothing like thatexists.
And then to you, that's yourskiff.
That's where you live out of.
And then you come out in theblazing Hawaii sun.
Speaker 01 (15:22):
Oh,
A'ndre Gonawela (15:23):
gee, yeah.
Shubhi Mishra (15:24):
you know, without any fans, but it's a ticky bar,
which is kind of cool, youknow, with like those
old-fashioned shades on them.
You come out, you know, youcatch up on life or you use the
unclassified version of acomputer to check on any updates
or how to fix a problem.
(15:44):
And then you go back in and dothe same thing.
And you do that day in and dayout.
And that is not easy.
But that's necessary to turnthat in into operational value.
And I think the last part, thedoing the do and the hard do is
where most of us fall shortbecause it's humanly, it's
(16:05):
humanly hard, physically hard.
A'ndre Gonawela (16:06):
No, yeah, exactly.
And I mean, like when we'retalking about, you know, what do
we actually do?
work on produce to gain thatoperational value?
I mean, what types of metricsor outcomes are we using to
determine if that capability istruly impactful versus being
more of that performativeinnovation?
I mean, I've spent a couple ofyears working in the defense
(16:27):
innovation space, and nowadays Isort of cringe when I hear the
word innovation being thrownaround by so many different
people who'll sort of use thatterm to just latch it onto
anything new.
So
Shubhi Mishra (16:40):
I think it starts with participating in
exercises, these exercises whichare replicating how a battle or
war may be fought.
And it starts with that, and itstarts with, can you do with
less humans?
Can you do with not awhiteboard?
How much can you automateend-to-end?
(17:02):
And the result of that ends upwith more and more operational
users, operators using yourtool, And then the word of mouth
circulating across the board.
So now that we have, let's say,a particular AOC tomorrow, we
have another one added.
And added because they want toand they want to use and they
want to make their life better,not because it's coming top
(17:24):
down.
And I think that is the metricthat we keep a close eye on.
And I highly recommend, youknow, businesses to do that.
And that requires a lot ofinvestment.
So instead of, you know, thisis, again, one of the
conversations I'm asked often,instead of investing in maybe a
large growth team or, you know,business development team,
(17:44):
invest in finding ways toparticipate in exercises and
really demonstrating valuebecause there's a wide gap right
now that exists.
And if you can make anybody'slife easier, they will not let
you go.
A'ndre Gonawela (18:00):
Oh, absolutely.
So I want to move theconversation into a little bit
about the work you do at Raft,because it's fascinating work.
I'm still trying to wrap myhead around it as I'm not a
technologist, so I'll definitelyappreciate your help in trying
to understand this.
But you talk about data fusion,autonomous data fusion, how
(18:21):
it's a big critical challenge intoday's defense environment.
Can you tell us a little bitmore about what autonomous data
fusion is?
Shubhi Mishra (18:30):
now data fusion is being done by human beings
using their gray matter and ifwe want to get to the level we
want to and compete with ouradversaries who by the way have
not only figured out end-to-endsystems and and can where
everything talks to everythingso it's just faster because of
(18:50):
the nature of it but also arecontinually investing in
technology um we must shift fromhumans gray matter to doing the
do to machines doing most of itautonomously and such that a
human is observing it and makingsure it's doing the right way.
A'ndre Gonawela (19:11):
So this is like the conversion of different
types of data into one universallanguage.
Shubhi Mishra (19:17):
So it extends a little bit beyond that, right?
So that is the data fusionprocess.
but autonomously requiresmachine learning models in an AI
to not only read what differentformats, standards, schema,
(19:37):
whatever you call it, whateveris coming in, and then translate
it back to how an operatorwould want it in how they
consume the information.
A, you know, a, somebody, a paycomp, somebody's area of
responsibility within a packhalf may be different from a
(19:57):
pack fleet.
And they, they consumeinformation different way, but
they need similar information.
These systems that they'reusing or consuming information
from are, have been built indifferent decades by different
vendors and, and, you know,which are not necessarily
talking to each other.
So it's always, how do you makesense of all this stuff into
(20:20):
something that the operator hasbeen trained on?
And the way different servicesgrow and teach their operators
very differently, and you haveto autonomously make sure the
information they're consumingmakes sense to them.
A'ndre Gonawela (20:37):
So it's like you have all of these different
capabilities, these differentplatforms, these different types
of technologies, computers,whatever, who are all outputting
all of these different types ofdata.
Say you have a team of people,one speaking Hindi, one speaking
Spanish, one speaking English,one speaking German, and you
(20:58):
have all of this great data, butthe one person who needs to
make a decision on it cannoteasily understand it, cannot
easily translate that into oneuniversal language like English
or something.
And we need to do thisautomatically.
So it's that sort of a veryrough allegory.
Shubhi Mishra (21:17):
Yeah, that's exactly it.
And what you said at the end issuper key.
It's like it is using theirlanguage, whether that is
English or something else.
And using natural language.
So it's not necessarily as ahuman as you would interact with
something.
And that's where the autonomouspart comes in, the natural
(21:38):
language part comes in.
So far, the tools that existand have been deployed are...
not natural language tools thatare not a human talking to
another human tool.
So imagine now, go back,imagine when operators are doing
their job.
When they pick up the phone andcall the other guy sitting near
the skiff, they're using humanlanguage.
(21:59):
They don't use machinelanguage.
But the tools right now thatexist, it's a machine language
talking to a human.
So if you want to change thebehavior and if you want to
change AI being the forcingfunction that enables them, we
gotta replicate this phone callto the upper operator and make
it frictionless and human talk.
(22:20):
So that's where the autonomouspart comes in and the data
fusion is just around themachine to machine talk.
A'ndre Gonawela (22:27):
Yeah, and I mean, when I'm thinking about
why that's so important, I'mthinking about all of the
different things that the DoD isprocuring, whether it's fighter
jets, whether it's sensors,whether it's the new unmanned
capabilities, whether it's allof these different types of
technologies, folks, that we'rethinking about that one may use
(22:47):
in a singular operation, youknow, when we're bringing one of
those capabilities in, we wantto integrate that into the
broader force to make sure itall sort of fits seamlessly.
And, you know, I'm noticingthis, you know, not just in a US
context, but like a lot of ourAsian partners, for example,
right?
They're like procuringdifferent types of capabilities,
(23:08):
right?
Like, you know, India is buyingfighter jets from France for
example, the Rafales.
Are those Rafale fighter jetsfitting into the broader
operational capabilities of theIndian Army?
Are they sort of making surethat all the data can sort of
talk to each other?
And, you know, I'm definitelyobserving that as we're
(23:28):
expanding the aperture ofacquisitions beyond the U.S.,
but into the allies, with theallies, especially with things
like AUKUS, with things like,you know, U.S.-India defense
cooperation and so on.
Is that...
Correct.
Shubhi Mishra (23:41):
You're 100% on point there.
And it's almost like you'resolving for the symptom.
symptom of new technology, newtechnology, but we are not
talking about the root cause.
And the root cause is, thereare only two ways to solve it.
One, the way China solved it.
It's one thing.
It's one company, onesubsidiaries of that company,
(24:03):
and they must talk to eachother.
And that's what I call aboutsolving end-to-end.
And then in America, wherethere's a multi-vendor ecosystem
and the best may win, we don'tfocus on solving this root
cause.
Yes, best may win.
Right now, humans are enablingthis talk track between all
these different systems.
What happens in the future whenseconds result in life and
(24:27):
death?
And that's not a conversationyet because we're still warming
up to it.
And I think that's the mostimportant conversation we need
to have.
A'ndre Gonawela (24:35):
No, yeah, absolutely.
Because I mean, you know, inthe current defense innovation
ecosystem, we still have a lotof work to do, but we have so
many interesting companies, somany interesting technologies,
so many new things that arepopulating, you know, our
capabilities.
But how are we connecting thatefficiently?
Because that's a key issuebecause, you know.
Shubhi Mishra (24:52):
It's the boring part, but the most important
part.
A'ndre Gonawela (24:55):
But it's vitally important, especially
when you're in the battlefield,when you have, you know, only
seconds to make a decision oflife and death, you need
processing power, right?
to automatically do this.
Because I mean, again, like,you know, folks, like going back
to the allegory about you havethe team of different people who
speak different languages.
If you have a human doing it,it's going to take so long to
actually do that translating byhand.
(25:16):
I mean, folks go back to yourFrench classes, right?
And think about how hard thatwas.
I mean, if you had a computerdoing it, it'd be fantastic.
So, I mean, you know, whenwe're talking about, you know,
universal data fusion, what doesthat actually mean?
look like in practice and whyis it so hard
Shubhi Mishra (25:36):
yeah i mean see it's hard because technically
it's hard right i mean to solvefor it's just decades of
different vendors decades ofdifferent systems and and the
way dod has wanted to solve forit to make a standard
standardized things and theproblem with that is and that's
(25:57):
why there's not a lot of buy-inand people who are making these
decisions don't understand theground truth, which is Space
Delta, Space Force Delta issolving for their specific area
of responsibility.
They will create something thathelps them get that information
(26:21):
faster and quicker.
On the other hand, you know,some organization within Air
Force or Navy, they will solvefor very specific ways they want
to solve for.
And so could they adopt thisstandard?
Yeah.
Would that result in themlosing precious time to enable
(26:41):
their decisions?
Absolutely.
And so this one standard, sizefit all, has not been adopted.
And I think it should be thereverse of it.
You do whatever you need to,but let's make sure the data
fusion autonomously happens.
So I think technically that'sone big challenge, just really
hard engineering.
The second is it isbureaucratically challenged and
(27:08):
there are a lot of policies thatstand in the way.
And those policies weredesigned for good reason.
And however, there are too manypolicies now and nobody wants
to share data with the otherentity because, I don't know.
To me, those reasons don't makeany sense, but they are
(27:32):
politically and bureaucraticallychallenges.
And I think a lot of those arealso fed by their different
vendors wanting to protect theterritory and making sure they
can protect the territory ifthey don't share data.
That's a huge problem.
piece of it.
And I also think the thirdpiece is, I don't think Congress
(27:53):
has looked at this problemstatement super closely.
I've spent a lot of time on theHill, continuously do so.
And I'm amazed at the gap theyhave between what the
information they know and gatherto what the information is on
the field.
And that's where, you know,they've appreciated my insight
(28:14):
because I'm able to translateand get, they can hear it from
the horse's mouth or almost fromthe horse's mouth of the gap.
And I think they haven't askedthese questions.
They haven't focused on this somuch and they just get pitched
how new things will solve thisproblem forever versus we must
(28:36):
find a way for the old and newto come together.
So more focus from Congresswould also help this.
But that's what I would say,the three challenges.
Technically bureaucratic andnot enough oversight,
congressional oversight.
A'ndre Gonawela (28:53):
Oh, yeah.
No, no, for sure.
And I mean, like, how do you...
I mean, when we're talkingabout how you actually do this,
like, how do you actuallyapproach integrating these
disparate data systems, youknow, from these multiple
vendors?
Like, what's the key to this?
Shubhi Mishra (29:07):
So...
Technically, there are thingsthat, at least the way we have
approached it, is reallydeveloping an abstract layer.
And so, you know, that givesyou 70, 80% of the way.
And then some of these thingsare...
Just similar type formats andso text format, you know, video
(29:32):
or pictures and just puttingthem in bigger buckets and then
figuring out how to parsethrough that.
So that's one way technicallywe have done it.
And the next is really a lot ofpartnership with other industry
partners.
And what that helps us is, andI would say any vendor helps is,
it fastens the getting to theend phase.
(29:53):
quicker and we get there fasterjust because then you don't
have the additional bureaucracyof just DoD and government
coming in the middle and you canjust make B2B connections.
And so their system can talk toyour system and, you know, you
get to the end result faster.
(30:13):
So that's how I would say wehave approached it.
A'ndre Gonawela (30:15):
No, for sure.
Definitely.
And, you know, we talked aboutthe information gaps belying our
congressional representatives.
We talked about, you know, whythere are gaps between the
forces and sort of thestylification of that data.
But do you think theacquisition community itself, I
mean, fully understands theimplications of fragmented data
(30:37):
architectures?
Shubhi Mishra (30:39):
I wish they do.
I wish they would.
They're not living the pain.
A'ndre Gonawela (30:44):
Why not?
Shubhi Mishra (30:45):
The pain must be felt.
And there's only one reason tofeel this pain is go out to
Guam, go out to Philippines andsee and experience a day in life
of that operator.
I think that will be gamechangers.
And I feel the ones acquisitionprofessionals that I've come
across who have in uniform, Ispecifically say that have
(31:09):
rotated out.
or have experience towards thefar edges are the most, provide
exponential results for thesecommunities because they
understand.
Yeah,
A'ndre Gonawela (31:22):
absolutely.
So I mean, When we're thinkingabout all of these challenges,
not just within autonomous datafusion, but sort of some of the
larger challenges, I mean, youallude to, I think, what you
just said, right, about theacquisition community, those
officials not reallyunderstanding the pain because
(31:42):
they don't feel the actual pain.
How do we work towards this?
I mean, what does the nextgeneration defense contractor
do?
look like?
How do they navigate thisintense, complicated, soulless
acquisitions framework with allof these different entities,
(32:04):
these different people you needto communicate the pain to, make
sure you're not communicatedwithin a silo and so on?
How do contractors navigatethis?
Shubhi Mishra (32:14):
So I think there is a now what we need to do, and
I think there'll be a differentfuture in a few years from now.
The now is unfortunately, wegot to bring the two communities
together and really have theusers advocate for what they
want.
And that's the other big gapI've seen is users don't know or
(32:38):
have a blueprint of how to evenask for what they want.
And that's where vendors likeus can help them because we know
how the game is being played.
I say game.
I have a lot of respect forthis game.
It's a beautiful game.
It's
A'ndre Gonawela (32:57):
a game nonetheless.
Shubhi Mishra (32:58):
Yes, yes.
But I do think it is educatingand enabling these operators on
the edge to go to theacquisition community to get
them what they need.
And it's continuously a lot ofconversations.
I also think the...
(33:19):
It is the only model we haveseen for success for any vendors
in this ecosystems are the bigprime or the legacy model.
And I think the new primesshould absolutely, and it's not
intuitive, it's against thehuman intuition, should
(33:41):
absolutely fight against thatand make sure that they don't
end up being bloated.
Because bureaucracy burns cash.
A'ndre Gonawela (34:20):
overly bureaucratic organization?
Shubhi Mishra (34:23):
Three letters.
ABC, always be cutting.
Cut what doesn't work rapidly,invest in what's working.
I think that's the simplest wayto structure it.
And the beauty of now withthese AI tools is you don't need
that much manpower, for lack ofa better word, to do a lot of
(34:47):
these things.
I think...
Invest in the people that wantto multiply and are 10x and not
invest in things and processesthat don't work.
A'ndre Gonawela (34:59):
Yeah.
And I mean, you know, I sort ofnoticed this with like a lot of
these companies that have beenstartups that sort of start off
small as they have success, asthey grow.
And you're in this sort of thisweird space where you're a
midsize company.
Growth is accelerating.
And then as a result, youalmost feel the need to put in
(35:20):
bureaucracy, right, to make surethat you can manage your
organization well.
I mean, people will often saythat being small and agile.
is the advantage.
However, how do you scale up anorganization to fill these
needs, to grow the success andso on?
Where's the middle ground interms of that management?
Shubhi Mishra (35:43):
So I'll tell you.
So it's one of those thingswhere, look, if a company is
selling that their capabilityand products is all around
enabling operators throughartificial intelligence, machine
learning, and data, they betterbe doing that on their side of
(36:04):
things, which means is they mustover-invest in these tools for
themselves and make 1x to 10x.
And so the same thing appliesbecause that's the premise of
everything.
And in my opinion, it's alwaysgoing to be, unintuitive because
(36:29):
it's so new, the capabilitiesthese AI is providing you.
But I think fight that urge tohave a bunch of middle managers.
Fight the urge for you need ahuman to help a human to help a
(36:49):
human.
Fight that.
That just slows everythingdown.
You have too many decisionmakers.
And I approach it verydifferently where it's who has
the veto versus the decision,right?
More confusion results stopsmoving fast and stops agileness.
(37:13):
And I understand this idea ofmiddle-sized company versus
small company versus largecompany.
It's measured by revenue, but Ithink a lot of that is
changing.
I do think in today's world,it's all about what value you're
creating and how you're solvingthe mission.
Thanks to not only the recentadvances in AI, but also the new
(37:39):
government in charge, they'revery much focusing on what
capability do you provide versushow large your team size is.
And given all of that, I reallythink it should be very much
focused on Solving the missionneeds, that creates a lot of
(37:59):
value versus, I don't know,scale with millions of dollars
that are humans that will bereplaced any day by AI in any
case.
A'ndre Gonawela (38:12):
Yeah.
And I mean, when we're talkingabout AI, I mean, you know, a
lot of people have a vision forhow AI can integrate into our
capabilities, but it's a veryvague vision blocked by a lot of
fog and a lack of clarity.
But, you know, there's also alot of caution in how AI
(38:32):
integrates into our defensecapabilities.
I mean, it should be...
does good AI adoption anddefense actually look like?
And what should it not looklike?
Shubhi Mishra (38:42):
So the way I think about it is just AI in
terms of decision-making lens.
As a human being, there are twotypes of decisions you end up
making.
As reversible, doesn't causelife and death.
Irreversible, your action ordecision will result in life or
death.
Majority of the decisions thatare being made within the
(39:03):
ecosystem are reversible.
You can change your decision ifit's not the right decision.
You can go back.
And I think that is absolutelyprime for anything artificial
intelligence brings in.
Where your decision causes lifeand death, we must proceed with
a lot of caution.
There's no coming back fromthere.
(39:23):
There we...
must test this.
There are a lot of exercises.
There are a lot of testing thatneeds to happen.
Even after that, there alwaysneeds to be an operator on the
loop or the human on the loop tomake sure the decision is the
right decision.
And furthermore, if you go intoit more, it's just not about AI
(39:45):
tools.
It's about AI tools that areshowing you how they're making
the decisions.
There are a lot of tools outthere right now in the
ecosystem.
You give them a question,they'll give you an answer.
That's not the right way toapproach it.
For a human and machine tobuild trust, the machine must
show how it's thinking and howit got to the answer.
(40:07):
And over time, you will buildtrust with it.
You like it or not like it, oryou tweak it.
But at the end of the day,showing that thinking is going
to be critical for thisintegration of decisions that
can be reversed.
A'ndre Gonawela (40:25):
Absolutely.
And I mean, how do we alsoensure the ethical,
operationally relevant andsecure AI integration in
national security missions?
Because I mean, that's some ofthe biggest concerns that people
will have about AI.
I,
Shubhi Mishra (40:40):
again, think of it as the two buckets of, you
know, if your decision can bereversed, then Let's weigh it
less in terms of these stagegates we have of something being
ethical or non-ethical.
Of course, nothing should benon-ethical, and that's why we
should be testing this out.
(41:00):
But over-emphasis on it,massive regulations on it when
it is an irreversible decisionand a life-or-death decision.
A'ndre Gonawela (41:09):
Gotcha.
So, I mean, you know, as wesort of run this interview out,
I mean, you've said fear is thegreatest enemy of progress.
I mean, how does that informyour own leadership style at
Raft?
Shubhi Mishra (41:20):
So I've always said run towards fear with a lot
of courage.
And I think as human beings,we, our monkey brain, our
amygdala brain, it's a fight andflight and the minute something
(41:41):
hard comes in, you just want toflee.
And I think that's the signalto run towards the hard problem,
run towards your fear with alot of courage.
And that is something we doevery single day.
And that is the sole reason whyraft is where it is and where
it stands.
(42:01):
And I, that is the sole reasonfor the curve of past six years
for us has been so exponentialthat we are sought out by our
customers to come help themversus the other way around.
And if you're not runningtowards fear with a lot of
(42:21):
courage, you're not having fun,right?
A'ndre Gonawela (42:23):
I mean, how does one have fun amidst all the
friction and the inertia ofthis defense bureaucracy?
Because I mean...
It sounds like a whole lot ofnot fun.
Shubhi Mishra (42:36):
It is.
Oh, the mundane parts areboring.
And yes, those are just thingsthat need to be done.
But it's so much fun to forge anew path and to redefine the
word prime, to really help theseoperators who need help.
There's so much...
(42:57):
Meaning, I mean, I'm going togo a little philosophical.
There's so much meaning in lifefor that.
I mean, you know, when youthink about all the time you
spent and all the nights youstayed up, and if you realize
that this is the impact one hadon the journey of a country that
(43:21):
impacts the world at thatscale, it's beautiful.
Couldn't be more fun.
A'ndre Gonawela (43:29):
So what changes, I think, technological
or institutional, are you mostexcited about over the next 12
to 18 months?
Because I mean, you know,within the span of like one
year, we see so manytechnological advances.
I mean, if we're looking atChina, we saw deep seek
suddenly.
come onto the world stage andthat set off a firestorm of
shock and awe and so many otherthings.
(43:51):
And, you know, we're in a newadministration now.
There are a lot ofinstitutional changes occurring.
What's exciting you the mostabout these types of changes
we're seeing in the short term?
Shubhi Mishra (44:03):
I will say, yes, there's a lot of conversation in
the media, in theadministration, on the hill
around AI, but it is not gettingto the edge yet.
It is not getting to the peoplethat can use it to make their
lives better.
And so I'm super excited to seethe adoption and the trust that
(44:29):
the operators start buildingwith these AI systems.
That's one.
And the second thing is there'sa lot of PowerPoint talk still,
a lot of innovation theater.
So, you know, that's the otherpart.
Now that the innovation theatercan act fast and can result in
action more because the time isnow, I'm excited to see those
(44:54):
PowerPoints being turned intovaluable tools that can change
lives.
A'ndre Gonawela (44:59):
No, absolutely.
Shubhi, thank you for joiningme here today.
This was a really greatconversation.
I really appreciate it.
I really appreciatedconnecting.
on, I think, artist Dane forinnovation theater, because I
think that's a huge blocker ofprogress.
But I mean, you know, I reallyfound the conversation on Data
Fusion really fascinating.
And for our audience members,definitely check out Raft's
(45:22):
work.
Data fusion is a very importanttopic if you're interested in
defense force modernization.
But also, I mean, think aboutShubhi's lessons on leadership
style and organization and soon.
But Shubhi, thanks for joiningme here today.
I really appreciate it.
t
Shubhi Mishra (45:40):
It was fantastic, Andre.
Thanks so much.
Thank you.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.