Ep. #31, A Nursery for Stars with Deepti Srivastava

Rachel Chalmers: Today, I am delighted to welcome Deepti Srivastava to the show. Deepti is the Founder and CEO of Snow Leopard AI, a platform that helps teams build AI apps using their live business data on demand.

Deepti has nearly two decades of experience in data platforms and infrastructure. As Head of Product at Observable, Deepti led the 0 to 1 product and GTM strategy in the crowded data analytics market.

Before that, she was founding PM for Google Spanner, growing it to thousands of internal customers, ads, PlayStore, Gmail, et cetera, before launching it externally as a seminal cloud database service.

Deepti started her career as a distributed systems engineer in the RAC database kernel at Oracle. She also serves as Vice-Chair of the Board of pbwc.org, a California non-profit.

She has a Masters in Computer Engineering from Carnegie Mellon. Deepti, thank you so much for coming on the show. It's great to have you.

Deepti Srivastava: It's great to be here, Rachel, thank you so much for having me.

Rachel: You started your storied career in databases, and you worked at the incredibly powerful and influential Spanner. How have those experiences shaped the way you think about enterprise systems and artificial intelligence?

Deepti: Yeah, first of all, thank you for that very kind introduction. It's interesting as I look back, like, sometimes I feel like I'm completely new to the tech world, given all the changes that are happening, and then I'm like, "Oh, wow, I've been around for a while."

And, yeah, all of that career has been in enterprise software, enterprise systems, you know, selling, building, evangelizing those things.

And so, to me, as I look back, I think there are two thing that really stand out in terms of what I really observed about enterprises, and enterprise application, and enterprise tech stacks.

One is, they're all unique ecosystems. You know, most enterprise applications exist in a context, they exist within their own ecosystem, their own tech stack, they were built, you know, at different times, and so there's sort of different ways in which the app stacks were put together that give unique emergent behaviors.

So, any new technology, whether it's, you know, cloud, or AI, or whatever databases, they all have to exist in that ecosystem, right? Like, having one awesome product or solution just isn't enough, right?

What I learned with Spanner, for example, is like Spanner is like, I still believe, one of the most amazing databases on the planet, but when we started actually externalizing it, the myriad of application stacks that existed from the biggest banks in the world to the smallest mom-and-pop shops, like, it was just so disparate, and each unique, and, you know, having just one product doesn't actually solve the problem for each of these unique systems.

Like, as we build as systems engineers, as infrastructure engineers, we think about how our system can, you know, conquer it all, and how our systems can be so beautiful and unique, and pure, like, each infra person wants to build the best system.

And I am one of those people, so I love that thought process. But when you start to really apply it to real-world scenarios, it's very humbling, because no matter how pure, and beautiful, and strong, and powerful, and globally distributed, and globally consistent your system is, it sits in somebody else's stack, and it has to work for their stack.

That I think is a truly humbling realization. And I think for me, one of the key things I hold as I move into, you know, building all the other products I built before, but also building Snow Leopard.

And I think the other thing that's just basically become a foundation of how I think is that data silos are here to exist, and they're reality of life.

Rachel: Mm-hmm.

Deepti: So, again, you can build really cool HTAP databases, lakehouses, warehouses, all sorts of houses, and all sorts of storage systems, and all sorts of LLMs and everything, but, like, this myth of putting all your data in one place and using it from that one place, it's just a myth. It's impossible to do.

Again, because each enterprise has different business units and different application stacks that were built at different times and have different technologies going on.

So there's no way that it's possible to consolidate data from across the org into one place, or through one pane of glass that you can just view it, one API to rule them all, right? Like, that is just not possible.

And because this dream is sold to enterprises, for good reason, because, like, it's hard to control and manage all these different disparate systems, like, there's so much data in movement all the time, which makes, frankly, developers lives hell.

Like, this is what I saw, because both with Oracle, which is, Oracle is still a huge install base for enterprises, and, again, with Spanner, both are very powerful and strong systems and hold crown jewels data for probably everybody in the world that is like in the enterprise space.

And, for all of them, it's just so difficult to manage that, and then to manage all the different data pipelines, data streams, ETL processes, reverse ETL processes, there's so much going on, and each of those things is so hard to maintain and so hard to manage, I saw this firsthand, and it really like stuck with me.

I'm like, "It doesn't matter what we're doing. Like, this unification platform for data is just really, really hard." And a decade ago, the SaaS revolution really promised that, right?

Like, the APIs for the Sola and the, like, microservices architecture was supposed to solve this problem, right? It was supposed to be like, "Oh, you don't have to worry about the data and the data silos, because the APIs are going to provide these clean interfaces through which you can communicate, they're going to be clean access control, and you can always, like, update and change things in the background, and the APIs will give you a clean interface."

And, frankly, that's just not proven out to be true, right? Like, APIs don't solve this disparate data problem across silos, across orgs. And as a result, eng teams have to build a lot of glue to maintain and manage these systems.

And sort of, this is why I started Snow Leopard in a way, because I wanted to help developers focus on building value for their customers and not managing and maintaining these data in motion and the APIs in microservices and all that sort of architecture along the way.

Like, there's a lot of glue to maintain, and, basically, all of your time and eng capacity and creativity, frankly, goes into just maintaining all these things. And I feel like that's such a waste of talent and, frankly, like, productivity.

Let's not do that, right? Like, let's leave people to build creative solutions for their businesses, for their users, and hope, like, "What if somebody else could take care of that gluing?"

Rachel: Yeah. Yeah, it's the classic problem of software. Like, "If there's 80% of somebody's job that's tedious and it's known to solve problem, why don't we automate it?"

So tell us what you're building at Snow Leopard.

Deepti: The aim for Snow Leopard is to bridge the gap between your AI systems, whether it's agents, or assistants, or chatbots, or whatever LLM-based applications, but to bridge the gap between that and operational business data in your org, right?

So if your chatbots and agents and assistants could really access sort of the crown jewels, as I keep saying, like, the business data that exists within your org, like, what if they could access it live and on-demand, right?

That would really unlock the potential of AI that everybody talks about, around productivity, around enhanced use cases, around just making our lives better, right?

A lot of that is still locked away in the traditional data systems and databases, data warehouses, and in API-based systems, right? And there's just no real way to connect them directly, right?

Again, there is pipelines, and data workflows, and stuff that needs to be both predefined and maintained, that just doesn't work in today's day and age, especially for gen AI applications, right?

So let me give you an example, maybe that illustrates this, if that makes sense.

Rachel: Of course.

Deepti: And this is actually one of those sort of things that ultimately really pushed me to start Snow Leopard. This is my own experience.

So, I was ordering flowers online for my parents for their anniversary, and they're in different country. And so I ordered flowers, and then, you know, at the last screen, you give your credit card and all that stuff, and then the final screen where you're supposed to get the order confirmation, it just didn't happen, it kind of just spun, right?

Rachel: Did it go through or not? Who knows?

Deepti: Yeah. So then the question is exactly that. And, really, what I want to know is not just whether my credit card was charged or not, 'cause I'll eventually find that out, the real issue is, has it gone through, or do I have to reorder the flowers, right?

Rachel: Yep.

Deepti: And I want to know that now, not like in a day or two days, right? When the credit card transaction is declined or accepted, whatever.

So what do you do? You go to support. And support is, generally, first-line of support is always chatbots. So you go to chatbot, and they expect five predefined questions, right?

So, I'm like, "Okay, what happened to my order?" They have the information of who I am, because of a cookie, and the browser, or something like that.

And so, they're like, the chatbot is, "Oh, what's your order number?" And I'm like, "Well, if I had the order number, I'll not be asking this question of what happened to my order, but clearly I don't have the order number."

So the only thing the chatbot can do at that point is transfer me to a human. They did, like, it's working as designed, they did that, and it took me 45 minutes to get to an agent who then looked at the Order Management System and said... You know, they have to look at two systems like your Order Management System and the confirmation system, et cetera, right?

So they're like, "Oh, yeah, the order went through it looks like, because there's a row in the Order Management System, but somehow the downstream workflow of sending you an email did not work. I see that the bit has not been set for email here, so let me push that email out, and you will get the order number."

So what does that mean? The data exists, and it shouldn't require a human to look at two screens. Like, the system has the capability today to be able to get that information and give you an answer in five seconds or less.

Rachel: Yep.

Deepti: But I had to sit on a phone for 45 minutes to do this. And I'm like, "Nope, there has to be a better way."

Rachel: That's a really good origin story. It's really simple and clear, and we've all had some version of that happen to us.

Deepti: Yeah. And, you know, the people will say that, "Oh, you could do this, you could build it this way, you could build it that way." And that's all true, right?

But I'm telling you just a very simple example, like, even today, with all the RAG-based gen AI systems, they're basically data pipelines that are ETL and data from all the systems, and putting it into another system, and then you ask that system for the answer.

But what does that mean? It's stale, it's not the freshest data, and it's probably not accurate. And so, what if we could just connect the system that is answering questions to the systems on-demand that have the right question, right?

So in Snow Leopard's case, what happens is, when you get the question, we look at the question and we realize, "Oh, you need to look at user information from Postgres," for example, "And order information from the Order Management System, so we can, in real-time, generate queries for both the systems in their native API, run that query on the systems, fetch the data, and then give the answer back."

All of this is happening in real-time, so the freshest data is being picked, it's happening on-demand, and it's happening in C2 it's happening in those systems, right?

So you don't have to predefine every single use case, and the data requirements, and the data flows for every single use case, right? But the way the technology is moving these days, so fast, and the use cases are moving so fast, and human beings like have ad hoc questions all the time about the stuff that they're doing.

So the old way of doing predefined workflows and predefined solutions just doesn't keep up for today's world. And that's the promise of Snow Leopard. let us build the glue of this data retrieval system, and you build the user value using your AI systems and your enterprise data.

Like, we'll just be in the business of fetching the right data at the right time from the right source without you having to build a bunch of, you know, spaghetti connections everywhere.

Rachel: It's a fantastic idea. You've been a Head of Product before, now you are founding your own company. How is that different, how is that leveling up? What did you learn that you didn't know before?

Deepti: Yeah, that's a really good way of putting it, leveling up. I didn't have a graduation ceremony for this, but, yeah, it's definitely leveling up.

I think the biggest thing I learned or continue to learn even today, 'cause it's early days for Snow Leopard still, is having courage of conviction in your idea, in yourself, and in what you're out to do, right? Whatever it is that you're aiming for, most people when they're building a company have some impact in mind, I hope. And so, having that courage of conviction is really, really, really important to be able to continue moving along, right?

So what do I mean by that? I think as a Product Manager, as an Executive, as a Tech Leader, you've most likely come across various things that you've either had to do or a case team, like, sales go to market, you know, messaging, content, solutions, architecting, and things like that.

So it's not that you're not familiar with all of those things, but when you're building a startup, you're inherently in an ambiguous environment where you don't actually know quite how to get there, right?

You just have to kind of go by instinct or listen to whatever people that have come before you, or, you know, your investors, if you have any, or, you know, use your network, or books, whatever it is.

But, ultimately, you're kind of in the game by yourself, or with your co-founders, if you have them, which, sadly, I don't have the luxury of co-founders, so I have to really do all of this myself.

But I think you get a lot of nos and you get a lot of pushback, and you get a lot of, like, just "Who are you and why should we believe you," right? And it's fair, right?

But, like, as a first time founder especially, I think having the sort of tenacity to get up every morning and be like, "No, I'm going to do this, no, I'm going to figure a way out, like, let me just charge ahead, because I know this is the right thing to do," et cetera, is I think really important.

Like, there is an art to listening to the market, to investors, to customers without being pulled in multiple directions, so that you can focus, right? Like focus is really important. And that focus comes from that sort of courage of conviction and making sure that you check in with yourself when you're like "This is the right thing to do."

Rachel: I have found a really clarifying experience. It's like a crucible, it's like, "Well, what do I absolutely know to be true? What am I completely convinced of that nobody else seems to see?"

I love that formulation as, you know, your specialization is the thing that's blindingly obvious to you, but not obvious to other people.

Deepti: Yeah.

Rachel: That's your essential weirdness that you can turn into your creation.

Deepti: I think the other thing that a lot of, especially first time founders, don't necessarily believe or know, is the sort of obsessing over customer value and creating, like, the thing you said, like what I see, like nobody else sees, like ultimately goes to, like, "I see this problem for my market, and this is the way I will solve it. Like, that's my expertise, that's how I know that this is the right thing to do, and the right problem to solve, and the right way to solve it."

Like that comes from being obsessed with your customers, right? Like, I say this, I've said this before, "Let me build cool stuff, and they will come," they being customers, is just a fallacy, right?

Like, I think I'm taking sort of a countercultural approach in the startup world with Snow Leopard of, like, I've always been obsessed with "What is the problem? What is the solution? And can I bring value to that solution?"

Like, "Do I have something where I see where the market is, how it's going, and what the solution for the problem can be?" I think that is really, I believe, fundamentally key to long-term success for a startup.

Like, building for your audience, for your market, for your ICP from day one is key to long-term success for a startup. Now, I'm yet to be proven right or wrong.

Rachel: It's a little crushing to hear that described as counterculture just because, like, all investors and all accelerators try to teach that to startup founders, but you're right. It's often, our advice is falling on deaf ears.

But you have to start with the customer problem. You know, the customer never buys vitamins, they always buy painkillers. You have to start with solving one of the top three annoyances that they encounter in their day.

Otherwise, why would they go through all of the effort and all of that learning curve and suffer all of the friction of introducing a new tool? That's the last thing anybody wants to do. The instigating incident has to be powerful enough to overcome that inertia.

Deepti: Yeah. I have a strong network of second time founders too, and, you know, first time founders, and technologists, and I see that difference.

And I think that's why I said this, because that was my instinct as a product person anyway, but to hear second time founders talk about that actually like reinforced it.

So hearing you as an investor, you know, and you invest in like earlier stage people when they're still figuring this out, right? They're figuring out PMF, and ICP, and things like that.

So it's sort of validating, again, to hear you say that, because that's just been my anecdotal experience. That's not the right way to do it, but also that most people are doing it that way.

Rachel: And your experience is pretty broad. You're an advisor to lots of other great startups. What are some cool things that you're seeing? What are some ventures that you're excited about right now?

Deepti: So I tend to invest in, or advise, data companies like shock and surprise, right? And, you know, I will be excited about a bunch of, basically, data storage system, like databases that are coming about that are actually more specialized, right?

Like databases are becoming more specialized versus more generic, right? And that sort of goes back to the data silos situation that I was talking about. It's only going to get more exacerbated.

But on the AI side, I mean, there's still a lot of hype, but I see a couple of interesting and cool technologies coming up, maybe others have seen this too.

But obviously document extraction has become like a defacto thing, but document extraction and specifically creating structure out of unstructured data is sort of this emerging thing, which I find pretty interesting.

Traversing knowledge graphs using LLMs and agents, another interesting thing, 'cause, you know, that will have impact on anything from sort of social networks and social media all the way to sort of recommendation systems and things like that, in a, again, ad hoc way without having to pre-train.

So I think that's very interesting. But, truly, I'm really excited about LLMs that are being built for non-English languages.

Rachel: Hmm.

Deepti: So that, I think is really exciting, because it democratizes this technology for the billions of people who don't speak English, right? That to me could be a total game changer globally.

I still believe that it's early days, like the best is yet to come when it comes to AI applications and AI technologies.

Rachel: And this really comes back to what they were built for in the first place. They started out as translation tools. But I totally agree.

I think if we think of the English language LLMs as, you know, fuzzy pictures of how we as humans sort and interpret the world, adding other languages adds entirely new ontologies, entirely different ways of encoding experience.

It's going to add much more resolution to that, what is right now a very fuzzy picture of our experience. I'm very excited about it.

Deepti: Yeah, truly.

Rachel: We are in the middle of yet another dramatic platform shift with these transformers, these LLMs. How do you think CEOs and other executives should be thinking about generative AI?

Should they just fire their entire workforce and replace them with agents?

Deepti: Yeah. I mean, I have said this before, I'm going to say it again, I'm less worried about AGI and the Terminator and more worried about "Where's my order," like we talked about.

Rachel: Mm-hmm.

Deepti: But, to me, again, I come from enterprise, I have a very sort of pragmatic view to technology into the world, so take my approach with a grain of salt. But, to me, we approach AI the same way we approach any new tool or any new technology, right?

Like, LLMs to me, or generative AI, are just another tool in the toolbox, and we have to figure out how to make it fit in our existing environment. Again, I am not a proponent of burn the whole thing to the ground and start with a new stack.

Like, I just, practically speaking in my 20 plus years, like, that's just never how anything has happened, even when you're talking about cloud transformations, digital transformations, right?

Like, they're never sort of, "Let's scratch everything completely and rebuild in the cloud." Like, that just doesn't work.

Rachel: It's all still running on Unix, 50-year-old software.

Deepti: Yeah. So I think the approach needs to be pragmatic and not just being carried away with hype blindly, frankly. And what are the key things we do with a new technology, right?

First, we have to assess the need. So, like, you can't just throw AI at everything, it's not the kitchen sink, right?

You have to see where you really need it, where it's really important, and, to your point earlier, like, the first top one, two, three pain points, where you can truly apply this technology to mitigate, if not eliminate those pain points. Like, that's one thing.

The other thing is the age old question of build versus buy. Like, do you need to build everything from scratch or can you buy it? And, at first pass, most people want to build, because, you know, it's cool, it's fun, you can like really control it. but, honestly, like, buying also means that you have to build around it, right? 'Cause, again, it has to be working within your environment.

So really being judicious about when you build and what you build is really important, because these things, any system, I think, frankly, but any new system that you don't even know how it works, takes a while to adopt.

And so, trying to like adopt it sort of wholeheartedly and building it from scratch, like it's just swallowing the whole cookie, and it's just not possible. So that would be the second thing.

And I think the third thing to me is like evaluating technologies quickly, which I think is exacerbated in the gen AI case, because technology is moving so fast, even I have a hard time keeping up with it.

So, having clear use cases, clear understanding of what you're applying and where, so that you can see value quickly, and then evaluating it quickly, is really important.

And I think this, again, is exacerbated for enterprises because most of the new tech or solutions are being built by smaller companies. And so, you really have to learn how to interface with them.

Ideally, you have clear ways of working with them and whether it's startups or small, you know, consulting firms, like, they should engage with you really clearly on really clear use cases, and, you know, provide you with clear milestones for how they're assessing value for you, right? And you can like sign onto that.

So being pragmatic about it and having a clear way of understanding how you're going to show value for yourself, for your company, is truly important in this day and age, more so than before.

Rachel: Really good news for those rare enterprises that are capable of establishing clearer success metrics upfront, challenging news for the rest of us.

Deepti: Yeah, yeah. And this is one of those places where I think we'll have to adapt to the new world quickly.

Rachel: Speaking of, what risks do you see in the widespread adoption of gen AI, and how might we mitigate them?

Deepti: Wow, there are so many thoughts in my head around this, but let me break it down into how I bucket it, which is technology risks, and, in this case, are ethical risks, or social risks, rather, of use of AI.

Rachel: Mm-hmm.

Deepti: On the technical side, like I said, you have to sort to the hype and really find real use cases and tools that provide value for you, your team, you know, your business and your customers.

Rachel: There's going to be a lot of wasted spend, we just dunno where it is yet.

Deepti: Yeah, exactly. And some of it is fine, 'cause we're still experimenting, but, again, going back to like the needs assessment, and like the sifting through the hype, right.

The second thing I think is like, which we just talked about too, I think scaling up your team and, you know, like learning how to build production AI systems is hard.

I think we're still at a stage where people who build production systems and people who build AI systems are not the same, right? They're coming together, but they're haven't quite come together.

Again, something that Snow Leopard's, like, the secret mission I have is, like, build this community of people who are systems people and also AI people, so that they can come together and exchange ideas, 'cause, frankly, the silos there are silly.

But, you know, understanding and training your team and training yourselves on like how to use LLMs, where to use them, when to use them, is pretty important. And then, you know, access to the right data for the right job, so that you can actually build value.

Like, I actually really care, obviously, you really care about the data access piece of it, because, you know, even if I'm super intelligent, or you're super intelligent as as human beings, like if you are given the wrong information, you're going to make the wrong decision, like point blank. And so, then blaming the technology for it is I don't think the right thing to do, right? Like you really actually have to empower your systems with the right information for the right job in order to really, truly see its value.

So I really do care deeply about that piece.

Rachel: Mm-hmm.

Deepti: That's why I think the technology side's at a high-level. The social side seems to be a bit thornier, in my opinion.

And I think, for me, if I want to just keep it to the top few things that are top of mind for me, I think the first thing is misinformation and sort of the use of AI, specifically generative AI for just gobs and gobs of misinformation that we've all experienced and seen, and has led to consequences already in our society.

That, and the use of AI as assistants, for example, is great. Like, replacing a human assistant with an AI assistant is perhaps like really useful and has real power with it.

But the reports that we hear, I'm sure you've heard them, about like kids having AI girlfriends or boyfriends, or significant others, and that causing them to self harm.

Like, I have a child, like I really, really worry about the uncontrolled and without guardrails aspect of AI in that form.

Rachel: It's fascinating, isn't it? Because these tools are a product of science and technology and engineering and math, and what's really needed to cope with them are critical thinking and textual analysis skills.

So, once again, the STEM people have created a problem that is best addressed by the weaponization of the humanities, ironically.

I agree, I see the same problem, and I feel incredibly, passionately that teaching people how to read, teaching people how to interact with these objects is incredibly important, and it's not easy to see ways to mitigate these risks.

Deepti: Yeah, I think that's the part. Like, you know, frankly, technology risks, we will figure it out, right? Like, these are systems, we built systems before, we built production systems before, right?

So, for example, I'm excited to bring my lived experience around building, you know, production enterprise systems to this world. But this is a new frontier for us as a society, right? When it comes to the sort of application of these technologies, right?

And like, frankly, it's just very early days. We don't know what kind of guardrails are necessary, even. Before applying any guardrails, we don't even know what guardrails are necessary. And so, to your point, I think we fall back to, or should fall back to, skilling up our, you know, society, especially younger people with critical thinking skills.

So I really appreciate you sort of saying it that way. The antidote for misinformation is critical thinking, and I think the antidote for this kind of thing is also critical thinking. So, yeah, I think that's a great, great point.

Rachel: And as we're talking about interacting with texts, what are some of your favorite sources for learning about artificial intelligence?

Deepti: You know, AI is such an interesting market right now, like, I would say for other types of technology, you know, there are standard places that you go to, whether it's published blogs, or news articles, or TechCrunch, whatever.

But with AI, everything is moving so fast that it's really hard to keep up with it, like, frankly. And, you know, for me, I tend to like have diverse news sources or information sources, right?

Like, from blogs, to LinkedIn posts, to archived articles, to podcasts, like this one, to, you know, just discussions among community forums, things like that.

I think you have to keep your ears and eyes open everywhere and all the time in order to really truly understand what's going on 'cause there's so much information out there, but also so much happening.

And, frankly, I think it's a really cool time because everybody is trying to share the information that they have, right? Anytime OpenAI comes up with something, I see some of my founder friends posting about it, right?

Like, you have the sort of OpenAI blog like it's VCs and founders and people who are just trying to understand the tech, and I have friends in the travel industry, they're coming out with things, right?

So it's really an exciting time, 'cause everybody's racing to inform everybody else. And, frankly, it benefits people like us, like, podcasts talk about this, like, you talk to people who are building, you know, first-seed companies, like where the first ideas are being formed around new technology that will change the world.

So I feel like there's so many of these types of information sources, and like people should diversify where they're getting the information from. Oh, and critical thinking.

Rachel: Yeah. I mean we make fun of the LLMs for hallucinating based on partial or incomplete data. We do the same thing all the time. And the better our data sources, the better our perspective is going to be.

Deepti: Yeah, totally.

Rachel: Deepti, if everything goes exactly the way you would like it to for the next five years, what does the future look like?

Deepti: Mass adoption of generative AI everywhere to make the world a better place.

Rachel: Mm-hmm.

Deepti: I truly believe this. Like, I believe in tech as an enabler for good. Like, I came into technology because I wanted it to have positive impact on people.

And I think we are at the cusp of, again, another sort of, as you've said, platform shift that can have huge impact on the way we operate and the way we really interact with the world, et cetera.

And so, I think gen AI is not just hype, even though there's a lot of hype right now, it's a game changer. It's just the game changers and platform shifts take 10 years, or multiple years anyway.

To me, I'm, as I've said before, less worried about Terminator and more worried about, "Hey, where's my order," having more efficiency in my workday, those kinds of things.

But for that to happen, people and businesses first have to understand, like sift through hype and understand what is actually useful and what is just hype. And then they have to be able to use these systems to help them make better decisions about what they're doing in real-time.

When you can use these systems to make real-time decisions, ubiquitously, for lots of different use cases, like, that's truly the adoption of this technology, and it become truly ubiquitous, just like cell phones, like, we can't operate without our cell phones anymore, for better or worse, that's a separate question.

But they truly have enhanced our lives in dramatic ways, but it's so embedded into our lives that we don't even see it anymore, right? And so, that's what I'm expecting-

Rachel: Yeah.

Deepti: For this to happen.

But, frankly, with current systems, like with RAG as RAG world closed, they're just not cutting it for the kinds of things that need to really happen. And new technologies and new solutions have to come into play that help you make better, more accurate decisions with these systems, right? And until that happens, we won't see the true power of AI.

But that's sort of why I'm excited about Snow Leopard and passionate about it, because I really want to help people make better decisions using their AI system and their own data, frankly. To actually make AI useful is the way I think about it.

Rachel: Two final questions on names. I haven't asked you this, why Snow Leopard?

Deepti: Ah, thank you for asking. First of all, 'cause they're really cute.

You know, when I was looking, to me, it was really important to have a name for the company that reflected the values I wanted to bring with the company.

And, as I was looking for it, I realized, animal names are really popular, but I wasn't really interested in beetles and stuff, I'm not an insect person. So as I looked up things, I was sitting and just browsing and looking at snow leopards, 'cause we'd gone to the zoo, you know, with my kid, and they'd seen snow leopards.

Rachel: Mm-hmm.

Deepti: And then I realized, like, when you look them up, snow leopards are actually, first of all extremely strong and agile, right?

Rachel: Mm-hmm.

Deepti: So like they all have power, but they're not in your face all the time. They're sort of reclusive, but will be there to do the things that you need them to do, which is the kind of qualities I want in the platform.

They're actually very good mothers, probably the best in the mammal kingdom. And, you know, raising early systems and like providing a platform for them to make them learn and things like that, like, they're again, qualities that I wanted the platform to have.

I want developers to build value and use creativity to build systems and not worry about the supporting infrastructure. And that's sort of what I thought Snow Leopard was, not in your face, but supportive infrastructure.

The third thing that was, you know, I'm a solo founder, or the third thing that was important to me, was that they come from the Indian sort of Himalayas and that region, and it's a little like nod where I come from and sort of what that means for me, and they made great stickers, so... Some of those will be coming your way later.

Rachel: Excellent. I will reciprocate with some "Generationship" stickers.

Deepti: Aw, thank you.

Rachel: A "generationship" is a starship that takes a journey that takes more than one human life to complete, so you have this multiple generations on the ship.

Deepti: Yeah.

Rachel: If you had such a colony ship to the stars, what would you call it?

Deepti: Nebula, which is a name I had thought about for Snow Leopard before Snow Leopard, but it's a nursery for stars, right?

And, like, it's the place where, I think, we're in that sort of stardust right now, right? Like, we're in a Nebula where new ideas are being formed, and nurtured, and they will change the way we think and operate.

And I'm so excited to be part of that, like, part of the AI revolution. Hopefully, I can help influence it and shape it.

But, like, I think Nebula speaks to that, like, being that sort of nursery for stars, generational nursery for stars.

Rachel: Absolutely beautiful. Deepti, it's been a delight having you on the show. Thank you so much.

Deepti: Thank you so much for having me. It was so fun to chat with you.