Building a Strong Data Ecosystem with Shubhi Mishra

[00:00:00] Shubhi: Users versus buyers is two different ecosystems and everybody, and because the wrong behavior gets rewarded, everybody focuses on these large conferences where the barge, uh, buyers are represented, or industry days where the buyers are represented and nobody is talking about the edges, where the users live and the 10 x improvements that they need in the solutions they're using.

[00:00:27] Shubhi: What their real pain points are. We choose to go to the moon in this decay and do the other things, not because they are easy, but because they are hard through our blood and our bonds. We crushed the Germans before he got here. You and I have a rendezvous with destiny.

[00:00:55] Bonnie: Okay, we're back at it again. I've got a rock star guest and, and to [00:01:00] be honest, I'm not even totally, I've never met her. I've talked to her one time, but so many people described her to me as a badass. It's just one of those things where you have to take the interview. So my name is Bonnie Evangelista. I'm with the Chief Digital and Artificial Intelligence Office, and I'm with Shubi from RAPT.

[00:01:19] Bonnie: Welcome. Thank you for joining me. 

[00:01:21] Shubhi: Thank you for having me and looking forward to this awesome conversation. 

[00:01:26] Bonnie: Um, tell me a little bit about who you are and, and try to bring, give us a little historical context of your working with government journey. Uh, 'cause that, that's kind of what we're gonna create some space for.

[00:01:37] Bonnie: And then I'll try to dig in where I can. But I, I wanna get, um, into the, how you became a badass part in this space. 

[00:01:45] Shubhi: Awesome. So, um, I'm a data engineer and a lawyer by training. Uh, I started rafting. Almost five years ago. Feels like yesterday though. It's been, it's been quite a journey. I'm an outsider. We all are outsiders in the majority of [00:02:00] raft, and what that means is we did not know what GovCon was until we found ourselves solving these hard problems for GovCon.

[00:02:10] Shubhi: And that's just. Started, uh, in 2018, and since then we have been growing and, you know, focus building that flywheel on solving those, those hard uh, problems. And to your other question about badass, I don't know, um, could be because I speak, uh, my mind and we take the, the path that is less traveled, um, and focus on doing the right.

[00:02:36] Shubhi: Thing, perhaps. Tell me a little bit more about Raft. 

[00:02:38] Bonnie: So you, you started, you had the, you have this data engineering background and you started this company. So what problem were you trying to solve with Raft? Yeah. Uh, 

[00:02:46] Shubhi: when we started Raft, uh, it was very simple, we found out as we were in the ecosystem, didn't see a lot of people using, for a lack of better word, their gray matter.

[00:02:59] Shubhi: [00:03:00] Uh, on the problem sets, uh, they, they were, they had in front of them, and it was very much focused on, uh, the lazy way of doing things and not thinking hard. Um, and, you know, in the ecosystem, we, we get rewarded for that. So even if you wanna break the, uh, mold and you wanna think differently, you don't because you're not rewarded for it.

[00:03:19] Shubhi: And so we saw this gap where things could be done faster, things could be done better. Things could be done in a way that accelerates the outcome. And we focused when I say things, we focus specifically on the data engineering layer that enables the machine learning layer, and right beneath is the infrastructure layer.

[00:03:40] Shubhi: So all that three stack, three layers of stack we focused on. And since then we have been building this flywheel that has been, you know, gathering a lot of momentum. Our journey within the defense started. With Cibber phase one that led to Cibber phase III and then, um, across different program offices, a program of [00:04:00] records, uh, for the different problems they have.

[00:04:03] Shubhi: Uh, one of our primary offerings is this data platform that is IL-VI, ready and operationalized. It's an edge-to-enterprise platform that we have built. So it works on the edge, but also. Can tackle the scale of the enterprise and it is differentiated with an output or a cop-agnostic layer. And it is data type agnostic.

[00:04:26] Shubhi: We push on no data standardization. It enriches data at the edge, as you transform at scale, it is schema-agnostic. So we never refuse anything as when the data on boards the way the schema is. So that is our core, core foundation. And then right above is our offering. The artificial intelligence layer or the machine learning layer, is on the tactical edge.

[00:04:49] Shubhi: With us, what we have is we, you're using limited language models and starting with the use case of battle management. And we are starting with that specifically around Air Force is [00:05:00] because that's where the data engineering's the most robust that we have seen all the bigs out there tackling this, solving this entire big elephant entire complex set.

[00:05:09] Shubhi: And we are starting the other way around, which is, hey, let's solve the smallest denominator. That's, uh, automate that, uh, add a machine learning layer where it makes sense, and then concatenate these solutions such that the entire problem set is solved. And as we are embarking on this, we have found that a lot of people or companies I should, uh, clients I should say, don't even have their infrastructure ready.

[00:05:31] Shubhi: And so. We, our third offering is a platform called, uh, platform Solution called Bidwell, which enables you to get to a software factory, for lack of a better term, in zero to 60, and then scale based on the solution and the needs you have for the 

[00:05:47] Bonnie: a platform that you offer. Uh, I've met with lots of companies that have similar platforms, so what makes yours?

[00:05:53] Bonnie: Different. 

[00:05:55] Shubhi: So there are two uniqueness in our, uh, platform that we offer the Bidwell. [00:06:00] First of all, it is agnostic of the mission area, so we enable scale. Of whether it be a cyber software factory or whether it may be an AOC supporting software factory, or whether it be a soft software factory, um, the, there's a layer of, uh, tweak con configurations that one needs to do.

[00:06:21] Shubhi: But the base layer of chassis, it's zero to 60 and we promise that and we deliver on it right away. And the second piece is that. Our platform is DDIL Ready, DDIL Standing for Denied, Disrupted, Intermittent, and Limited Impact Ready. And that is very unique given where we hear within the, um, defense ecosystem where everything is going and how anything we build needs to be.

[00:06:54] Shubhi: Be operating in a denied and disrupted environment. So those are the two unique [00:07:00] characteristics of the platform that we offer. 

[00:07:02] Bonnie: I heard you say earlier that it sounded like this was, how are you applying that No. Data standardization was kind of a, something you promote as part of what you do. Can you clarify that?

[00:07:13] Bonnie: Because, and, and then we can dig into it a little more, like what does that mean to you and what, what in your, in your context. 

[00:07:20] Shubhi: Yes, given that DoD is such a, what we call it, an open source ecosystem, where you have so many of the users with such disparate data types and data standards, that it's almost impossible to standardize it at the onset, at the receiving end of data.

[00:07:40] Shubhi: And so. The way we are proposing our solution in incorporate is to receive the data as it does not matter. Do the transformations enrichment at the layer where our data platform exists, and then provide it in a format that is consumable by the majority of whatever the [00:08:00] output may or whatever the cop may be.

[00:08:01] Shubhi: And what we have seen, the differentiation or the way it's been done wrong is usually the platforms specify the format, they will consume the data. Data, the raw data, which in our opinion is not the way that one can scale given the behemoth of DoD data. I 

[00:08:20] Bonnie: feel like that's an unpopular opinion depending on who you talk to.

[00:08:23] Bonnie: Is that your 

[00:08:23] Shubhi: experience as well? Absolutely, and I think it, the devil is always in the details and it's easy to say, standardize versus to say doesn't matter, consume it. And then figure it out such that you can feed whatever you need to feed as an output. It's, it's, uh, it's generally, it's, you know, not as just, just this part of it, but other parts of, um, a, what I call the legacy DoD mindset, um, is everything that works or the suggestions that they're given [00:09:00] advisor there given is why change, why do something upstream or flow against the, the.

[00:09:07] Shubhi: The stream when everybody's talking about data centralization. It is, of course, the way we have done it and we have failed. Why change it now? So yes, I think as a concept, uh, it is not popular because it requires a lot of hard work to get it right. Yeah. 

[00:09:22] Bonnie: Uh, you're not the first person. An industry that's said something similar.

[00:09:27] Bonnie: Uh, I've, I've yet to hear anybody on the government side start to embrace though what you're saying. So have you seen, I'm assuming success using this methodology 

[00:09:36] Shubhi: or this model? Yes. Um, right now we have operationalized this platform in two very different ecosystems. One is the ABMS within the C-III-BM office.

[00:09:51] Shubhi: And the other, which is a completely different use case, is for the SOF community, the SOCOM, um, data layer. And both of [00:10:00] them deal with different types of data they're consuming. Um, whether that be, um, you know, flight data or that may be intelligence report structured unstructured. Doesn't matter. It's disparate data, data types, and datasets.

[00:10:16] Shubhi: But it's the same core app platform that's being used across both the platforms, um, both the mission set threads, I should say. And we are continuing to build and they are continuing to use. Uh, different types of users are continuing to use this, uh, platform and its scaling, uh, to provide real-time data-driven decisions.

[00:10:38] Bonnie: I'm starting to see why people describe you as a badass. I don't know if, uh, you may not see it or not. I mean, you're, you. You are taking what people think is impossible and you're making it real. I mean, the, the, the, you just, just named two heavy hitters, I would say, um, in the government who are doing this work at, at the Edge.

[00:10:59] Bonnie: [00:11:00] And I do have a question. Since you said C-IIIBM, are you also making the data interoperable? Across and, and I don't know what their use case is, we don't have to get into specifics, but across either domains or use cases and whatnot, is that also part of what's happening at the data layer? 

[00:11:17] Shubhi: Yes. Uh, that is a core piece of not only a data platform, one of them, I.

[00:11:23] Shubhi: Prerequisites to solving the problems they have. And we are, we may be, we are very close to it, embarking on this journey where there is an overlap between some of the data that ABMS is consuming and some of the data that SOF wants to consume and vice versa. So we are working, uh, also towards bridging both of those agencies and bringing them together.

[00:11:50] Shubhi: Um, uh, because as you know, SOF operates in, you know, uh, their user base is super wide and they have different needs and data and [00:12:00] operability. That's my accent coming in the way, but interoperability is, is, is a primary key requirement for scale, 

[00:12:10] Bonnie: for for sure that in itself is the problem. Of the day, I would say, or the, or the year I suppose I'm curious in, in your journey, because you said in the beginning this, this was not your world.

[00:12:25] Bonnie: Like you, this whole GovCon environment, this wasn't your thing. So how are you finding the right users and the right buyers? To take you to where you are today as 

[00:12:35] Shubhi: we, and we made a lot of lessons, you know, mistakes, uh, as we went through our journey and learned from all of them such that we did not make those mistakes again.

[00:12:44] Shubhi: And one of the key initial mistakes was the advice that was given to us from all the industry veterans for the, um, I air quote them. On, Hey, it's all about relationships. It's all about knowing the person who knows the person. [00:13:00] And we tried that for a few, and it just was not working out in the way we wanted to work out, whether that be getting to solving the super hard problems, accelerating that.

[00:13:11] Shubhi: Um, and it was, it was just the, it was just slower than what we had anticipated. And sometimes it was in some situations just wrong. As we continue to embark on this journey, what we realized thankfully, um, sooner than, um, going down a path, uh, that would not be helpful for Aft, is the users versus buyers is a two different ecosystem and everybody.

[00:13:44] Shubhi: Because the wrong behavior gets rewarded. Everybody focuses on these large conferences where the barge, uh, buyers are represented, or industry days where the buyers are represented and nobody is talking about the edges, where the users [00:14:00] live and that 10 x improvements that they need in the solutions they're using and what their real pain points are.

[00:14:08] Shubhi: It's just assumed that a buyer will communicate those pain points, even though that is probably three or four degrees separated and filtered through, and a buyer rarely is a user, and that just gets lost in. And so in our experience going to the edge node, the edge is no edge, edge, most edge node air quote again.

[00:14:33] Shubhi: Is, has been a great differentiator and, and this, this includes across the organization. So I, um, barely been in the office. I'm always traveling, talking to the ones that are on the field, um, learning from what they have and what start working and then taking that information back. You know, seeing, see our role as in even informing the buyer, Hey, do you know about this?[00:15:00] 

[00:15:00] Shubhi: And then informing our team, do you know about this? Such that we are very creative in how we solve the problems for them. And it may not be in the scope of work, but that's fine. But that's a conversation of you realize that this is, this is where. The product roadmap needs to be iterated such that you are solving the problem for the user.

[00:15:20] Bonnie: I exactly know what you're talking about. I think most people who listen to this show know what you're talking about. So take it down or get a little bit deeper with us in how you navigated that to get to the edge and get that 

[00:15:34] Shubhi: feedback? Um, I'll give you an example. This is an example where we failed and thank goodness for we failed.

[00:15:40] Shubhi: So we had this Sibber. Uh, back in, I guess couple, a few years back called Feed one and we have navigated this server from of one to a two to a three. And this was one of those, you know, back in the day when everything was one, the mesh one, the feed one, the platform one, so this cloud one, so this was a [00:16:00] phase one server and we were so excited and we crushed delivering the product within the first six uh, weeks.

[00:16:09] Shubhi: What happened is we were talking, we were not talking, and at post six months they were like, oh, sorry, this is not it. We don't need your help. Um, good luck and goodbye. And we were baffled like what happened here? And through that experience, we learned that a user and a buyer are a very different community.

[00:16:28] Shubhi: And at the onset of A. Program like this, you gotta engage the boat, unfortunately, you also make them talk. And uh, fortunately, or unfortunately, it's part of a job to bridge that gap. And, and since then, that is the playbook we have followed. So we start that at the onset and it's a continuous, um, conversation and place we have in any of the programs that are non, uh, typical, I would say that we follow and.

[00:16:59] Shubhi: [00:17:00] That continuous engagement has led to success and folks even discovering and buyer user community, both discovering what the gap is. Um, because sometimes a user may want your product, but they don't understand how it can be bought. And so the, you know, the, the, as a vendor. The more coaching you can do to bridge that gap works out to one's benefit.

[00:17:26] Shubhi: You, you're 

[00:17:26] Bonnie: one of the few people in the industry I've heard take responsibility for the challenge you just described where you, you become the bridge even though that might not be fair. To ask you to do that. It's not maybe your job, um, but you're kind of incentivized to do that to so that you can sell your product and you can bring solutions to the department.

[00:17:48] Bonnie: Uh, I'm just shedding or maybe putting more color on the why you're a badass theme because you, you are, uh, not on trend. And in a good way though. In a good [00:18:00] way. That is, um, that has promise, I would say. So with that, you are so, you've, you've, you understand the nuance I think a lot of people take a long time to get to and to understand.

[00:18:10] Bonnie: Um, I also heard have, you know, you and I, when we were conversing you called the Valley of Death Moronic. And I think this ties to what you're talking about. So can you elaborate a little bit more about it? Your thoughts and feelings on the valley of death because the idea is because I, I believe if you do what you're saying, there is no valley of death, but 

[00:18:29] Shubhi: go for it.

[00:18:31] Shubhi: Thank you. Um, it is moronic and I sometimes get in a lot of trouble for this, but it's the truth. Uh, and it is, a moronic term for me because it tells you the outcome. Ends up scaring some of the outsiders before they even, or even insiders. Um, and again, air quotes, uh, or, or the ones that have been in the, you know, system for a long time ends up scaring them and telling them the outcome before there are even fundamental ways of rethinking how [00:19:00] to do this.

[00:19:00] Shubhi: It is. I don't, I don't, I think the value of that is a super awful term. I think the term should be perhaps the value of product market fit, and the focus needs to be on finding, making not only that product 10 times better than what currently exists, but also figuring out how to get it. Um, the user, the customer on the other end to buy into it and to say yes.

[00:19:26] Shubhi: And, in my opinion, what happens is the SIPR one and phase the SIPR two have a different method and approach to getting them to getting the signature on, on the line, saying, yes, you've been awarded. And, and then beyond that is a completely different approach. So what I see is people and companies are not able to pivot, are not able to.

[00:19:50] Shubhi: To rethink their strategy because they don't understand it is a different beast, and we don't do, as a larger community, we don't do a [00:20:00] service, uh, to ourselves by calling a valley of that, which assumes no matter what you do, you're gonna die. And so, yes, I, I mean, if it wasn't for the Sibbers, we would not be here.

[00:20:13] Shubhi: So we are very thankful for it. And anybody, you know, who's. Embarking on this journey just needs to think. It is a completely different beast. And the approach, the people, the team, all need to pivot and change for the, uh, assuming the product is good, of course, but there is a user, if your product is good, that needs it, and it's your job as a vendor community to figure that piece out.

[00:20:35] Shubhi: How did you 

[00:20:36] Bonnie: differentiate the user from the buyer? I, I know you've already had some narrative on that, but I'm still, um, trying to place for others out there. Maybe there's a nuance there that might be helpful to them. 

[00:20:49] Shubhi: Absolutely. So. In this, uh, example of let's say Air Force Community, the user may be a fighter pilot.

[00:20:57] Shubhi: The user may be an air battle [00:21:00] manager, or C-II, but the buyer is ACC. The buyer is AFLMC. Those guys don't talk as often. They should be because they're not, for whatever reason, they've not been located, they don't, they have three degrees of separation. But you know, the ABMS office has done, uh, see the C-IIIBM office has done a good job of bridging that gap.

[00:21:22] Shubhi: Largely it is disjointed and what we have seen as the user, the operator, they are, um, located in these, um, for the lack of a better word, some sort of remote base out there. And they're not there making decisions on what gets funded and what doesn't get funded and why it's important. Um, but it's, it's just been through an experience that we have been able to determine, the difference.

[00:21:47] Shubhi: And I think the other ALS also part is. Um, and this is again, the fly we've, we, we've been building on. So for us, it's obvious and easy to see, um, because there were so many failures at the beginning of Raft is, so if we are [00:22:00] reactive, so which that means, hey, just submitting a response to an RFP that comes out, then, it doesn't matter.

[00:22:08] Shubhi: User versus buyer, then your buyer is what you need to. Um, respond your solution way, then, you know, sure, you can sprinkle some thoughts and ideas, how you can help the user, but you need to write it in, uh, how the buyer wants to buy. But if you're shifting left and you are, if the solution is going to solve or create a 10 x increment in value, then it does start from the user.

[00:22:33] Shubhi: And a buyer. And then, you know, bridging that gap such that a fighter pilot gets what's its need and ACC has requirements that it can put the money against. What, what's a pro tip 

[00:22:45] Bonnie: or, um, your, a best practice that you would encourage others to follow for bridging those two communities together? End user and buyer.

[00:22:54] Shubhi: One of the things that's been super helpful for us is being part of those. Air quotes again, [00:23:00] underground conferences that happen where you do see a lot of the users go and some buyers go, the buyers who care go there. And I think that's been, those kinds of conferences are not popular, but those con conferences, um, or events, they probably are not conferences 'cause they're not huge.

[00:23:19] Shubhi: But yeah, events are super helpful. And, you know, and the, uh, a buyer who cares if you invite them, they will go. And so that has been a great area for us to connect the two. And then a lot of this is unfortunately, you know, pounding the pavement. It is that continuous conversations with the edges on both sides and finding the buyers that care because they will have that conversation.

[00:23:41] Shubhi: They will come out and meet you at these different Air Force bases or whatever. Um. Military bases rather wherever, you know, the users are. So that's been helpful. Um, and then post COVID, everybody can, you know, come on teams and, and have those conversations. Um, but there [00:24:00] are, there's a subculture that exists within the communities that Care GSA is a great, uh, way to bridge that gap.

[00:24:07] Shubhi: Their cyber phase. Threes have been phenomenal because they're a different kind of buyer and they understand and move through the acquisition process super well, and, and so all these mini things, there's not a playbook. It's, it's like a bunch of little things that need to come together to create magic.

[00:24:22] Shubhi: Everyone uses 

[00:24:23] Bonnie: the term, it's a team sport. I like to think of it as puzzle pieces and someone's gotta put the puzzle pieces together. Uh, 'cause it doesn't always happen very naturally or organically. These are things I've talked about with other guests, but you're adding a different lens to it. It's, it's been nice.

[00:24:40] Bonnie: I'm gonna bring it back to where we started with where you're talking about good data foundations and that's kind of what, what your company provides. So what would you encourage the government in terms of how to build a strong data foundation? So there are lots of buyers in the government, lots of end users.

[00:24:57] Bonnie: What? What can we 

[00:24:58] Shubhi: say to them? Resist [00:25:00] the urge to make a one-size-fits-all solution. It is counterintuitive, but that is exactly what's needed. Given that I. The government is not a Netflix or Uber, where you control the end-to-end. Make sure the data foundation is strong before the Data Engineering Foundation is strong.

[00:25:20] Shubhi: Before we start talking about the machine learning layer, there's a reason, even though we are. Spread across the different agencies, um, and different mission threads. Very few of them use our machine-learning tools. And the only ones that can use it at scale are the ones that have a very strong data engineering foundation.

[00:25:38] Shubhi: So let's not talk about machine learning. Let's talk about data engineering first. The third thing is to be creative when it comes to thinking outside the box regarding the scope of work and what outcome. You're seeking vendors, outsiders want help. And that's why we are here. So if you just tell us the end [00:26:00] result, we will get you there.

[00:26:01] Shubhi: But there has to be trust on both sides. And I think that is, I. An assessment that happens over time. But if there are ways, including questions like how did you fail in the last time you wrote some, a response or a program that you happened, and people can tell you the truth. Uh, maybe those are the ones you wanna do business with.

[00:26:24] Bonnie: Those are some stellar recommendations. I wanna take a little layer deeper into the Strong Data Engineering Foundation. Can you give an example of what that doesn't look like? 

[00:26:34] Shubhi: So someone that does not have strong data engineering, and again, it's defined on the exactly what the. Use case. The need is a business use case will be very different from a war-fighting use case because the low latency data requirements or real-time requirements will be very different.

[00:26:51] Shubhi: So it depends on the use case problems trying to get solved, but if it's not a good one, it will break when it scales. Absolutely will and scale [00:27:00] and data means different data types. Scale and data mean in government specifically cross-domain, real-time streaming scale, and data mean the use of simulated data for machine learning versus real data.

[00:27:14] Shubhi: And use of predefined schemas and not necessarily schema on the fly. So all those things are necessary for scale, for real, real-time transfer across different security classifications that if it doesn't exist. There's no way on the other side, one can use machine learning to teach these models with real data and we'll continue to see simulated data being used and then these models not working when in time of need, my organization 

[00:27:45] Bonnie: this is the CDAO.

[00:27:47] Bonnie: Chief Digital and Artificial Intelligence Office, and Dr. Martel, who is leading our organization, he talks a lot about quality data and he says that's the key to machine learning, ai, all the rest of it. What [00:28:00] does quality data mean to you? Because I, we already talked about, okay, maybe we shouldn't be overthinking data labeling earlier, but so if we're not doing that, what does that look like?

[00:28:09] Shubhi: It's, uh, quality data is so, um, subjective and that's a great question. I would go back to the raw data here because quality data in our opinion, is in the moment or in the moment when the attributes for quality are defined. And that would be asked for the x mission thread. It would be different for different mission threads.

[00:28:33] Shubhi: And what needs to be across the board is a way to preserve. The raw data is such that when the schema changes such that when the downstream stuff changes, the raw data can deliver on the quality that is defined in that moment, or for the need to make whatever decisions you need to on the right side of things.

[00:28:55] Shubhi: I 

[00:28:55] Bonnie: mean, what I'm hearing the most from a lot. Of this conversation [00:29:00] is I identification of problems and gaps, mission gaps first, and then the, the solution really from your perspective, can come pretty easily if you're just very clear and, and open and transparent about that. So that, that is, that is something we're trying to promote on our side and, and I.

[00:29:16] Bonnie: Appreciate that, that perspective of yours. What else do you wanna leave us with that maybe was left unsaid? You wanna drop any more, any more bombs on us in terms of where we need to be rethinking 

[00:29:28] Shubhi: the status quo? I think, um, while everything is long. Um, view of things, which is fantastic, I think. Um, and I think, you know, the subcultures, including your organizations, have done a great job of this, of resisting, uh, a forever solution and rather focusing on what is the need now.

[00:29:49] Shubhi: I think that is something we, we need to continue. To think about and pivot, uh, from when it doesn't make sense or evolve. I think it's an evolution. It's not a change in strategy. I think we need to [00:30:00] continue, uh, more down that path. And this is again, a counterpoint here, but we, this notion that we have within the defense that somehow the commercial or the private companies are going to come and just have a solution that is.

[00:30:17] Shubhi: Plug and play. I, that's a very interesting comment to me when it's made, and I find it hard to believe anything is plug and play because of the nuances that exist in defense. And from my vantage point, it is, I think the flywheel of the. Not only the business model but also the offering of the nuances until the 80% mark makes sense.

[00:30:40] Shubhi: But the remaining 20 to get to the full solution are very much dependent on the environment that the technology supports. So I think we need. To again, think of this new way of thinking where it's not, Hey, private companies are awesome, or commercial companies are awesome. Um, let's just private sector companies, I [00:31:00] should say non-DOD companies are awesome and, um, let's let them just come and solve for us.

[00:31:04] Shubhi: It's, it's much more of, they, do they. 80%, which is great, but the 20 is very, very specific to DoD. And then how many of these companies are building their flywheel on perfecting that 20% versus getting us to adjust to an 80% solution because that doesn't solve the problem? 

[00:31:23] Bonnie: This was, a ton of fun, and I learned a lot from you.

[00:31:26] Bonnie: I can't thank you enough for sharing some time with me. 

[00:31:28] Shubhi: Thank you so much. This was such a fun conversation and very much of a conversational format, which it doesn't have. Rarely, which rarely happens, so appreciate your time.