EP 121: Snowflake CEO Sridhar Ramaswamy on the Future of AI and Data

Nine months into his role as Snowflake’s CEO, Sridhar Ramaswamy sat down for a candid discussion on the journey so far & what’s ahead. He uncovered details about his transition to Snowflake CEO, the operational changes he’s put in place to accelerate product development, and his bold predictions for AI’s future—along with how he’s guiding a vision for Snowflake to fit into that landscape. Sridhar also reflects on how his leadership has evolved, from leading a major Google unit to founding Neeva (acquired by Salesforce) and now running Snowflake. [0:00] Intro [0:56] Sridhar’s Journey to Snowflake [3:37] Snowflake's AI Vision and Cortex AI [10:35] Innovative Uses of AI at Snowflake [15:49] Challenges and Future of AI in Business [21:27] Snowflake's Long-Term AI Aspirations [25:04] Predictable and Unpredictable Aspects of AI [26:20] Autonomous Agents and AI Predictions [27:01] Market Consolidation in AI [29:54] Journey into AI and Early Experiences [31:21] The Evolution of AI Models [32:06] Building Neeva and AI Search Engines [36:16] Leadership and Product Development at Snowflake [40:45] Setting Goals and Accountability [43:25] Leadership Style and Cultural Foundations [47:25] Career Advice and Finding Your Path Executive Producer: Rashad Assir Producer: Leah Clapper Mixing and editing: Justin Hrabovsky Check out Unsupervised Learning, Redpoint's AI Podcast: https://www.youtube.com/@UCUl-s_Vp-Kkk_XVyDylNwLA 🎥 Subscribe on YouTube: https://www.youtube.com/channel/UCugS0jD5IAdoqzjaNYzns7w?sub_confirmation=1 Follow on Socials 📸 Instagram - https://www.instagram.com/theloganbartlettshow 📱 X - https://twitter.com/loganbartshow 🎬 Clips on TikTok - https://www.tiktok.com/@theloganbartlettshow About the Show Logan Bartlett is a Software Investor at Redpoint Ventures - a Silicon Valley-based VC with $6B AUM and investments in Snowflake, DraftKings, Twilio, and Netflix. In each episode of The Logan Bartlett Show, we sit down with the people behind today’s most important startups and extract the tactics, lessons, and frameworks they’ve learned the hard way. Conversations span hiring to GTM, product, growth, fundraising and everything in between - collectively forming the ultimate playbook to make you a better CEO, investor or board member. Tap follow and enable notifications to stay ahead of the game.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:03Speaker 00

  Welcome to the Logan Bartlett Show. On this episode, what you're going to hear is a conversation I have with CEO of Snowflake, Tridhar Ramaswamy. Tridhar and I talk about a number of different things, including his transition to leadership at Snowflake, how he's been able to make operational changes to the business to help them execute on shipping product faster, his predictions for the future of AI, and where Snowflake sits in that world, as well as how his leadership style has changed from the early days of Google, where he led a major business unit, to Neva, the company that he co-founded, as well as today with Snowflake. A really fun conversation with a number of different predictions about how AI is going to play out in the future that you'll hear now. Thanks for hosting. This is a

• 0:46Speaker 01

  Of course.

• 0:47Speaker 00

  It's a cool room. We're in the, what do you call this, the gaming room?

• 0:50Speaker 01

  This is the game room.

• 0:51Speaker 00

  Yeah, game room turned podcast studio. It's a hybrid. You don't typically see that. So you've been on board now as CEO for 10 months?

• 1:00Speaker 01

  10 months, roughly.

• 1:01Speaker 00

  And how long at Snowflake?

• 1:02Speaker 01

  Let's see, joined the day after Memorial Day last year, so like 17, 18 months.

• 1:07Speaker 00

  17, 18 months.

• 1:08Speaker 01

  But, you know, it's especially because you're inside, it's a little bit of an awkward process in that it's not like a formal offer letter, and so you're sort of like waiting.

• 1:23Speaker 00

  Guys, are we doing this? When are we doing this?

• 1:25Speaker 01

  Yeah, when are we doing this? And I also had a few other opportunities that I had to like go say no to. And, you know, opportunities don't come all that often. And so saying no to some concrete things while saying this will likely work out, super awkward.

• 1:48Speaker 00

  So, in the original plan, so Snowflake acquired your company, and your original plan was to stay on for

• 1:55Speaker 01

  Six months.

• 1:55Speaker 00

  Six months.

• 1:56Speaker 01

  That was my deal with Frank, that I would stay on till the beginning of December, and then go figure out what I was going to do with my life.

• 2:02Speaker 00

  Was it at the six-month mark that you made the decision to hang on a little bit longer than that, or was it somewhere earlier than that?

• 2:08Speaker 01

  I started looking, I would say, I started talking to a few folks, October-ish, last year. These things take time. Senior roles don't come by that, you know, that often.

• 2:20Speaker 00

  What was your initial mandate in that six-month period of time?

• 2:25Speaker 01

  I was SVP of AI during those during those six months. And I joke to people that I was like a cabinet position without an actual portfolio.

• 2:40Logan Bartlett?

  You were kind of an IC, right? Honestly.

• 2:42Speaker 01

  Exactly.

• 2:42Logan Bartlett?

  No direct reports.

• 2:43Speaker 01

  No, no direct reports. But roughly the mission that we set for ourselves was to make sure that Snowflake was kind of an A player when it came to data and AI. It was really to craft that product vision along with Vivek, along with others that came like Baruch and Asim, about how do we bring AI front and center to Snowflake in a way that stayed true to the product principles of Snowflake. One product, everything is tightly integrated, stuff just works. And so that was the bar that we set for ourselves. And so it was a lot of product design.

• 3:29Speaker 01

  And then there was also a lot of like talking to customers about AI, what was coming. So that was roughly what I did.

• 3:37Logan Bartlett?

  And so now, I guess you're 18 months into this AI journey within within Snowflake. I'm curious, how would you describe where Snowflake is in AI today?

• 3:49Speaker 01

  First of all, I think of AI as a as a horizontal capability with broad application. It's a little bit similar in that sense to, let's say, like the mobile change, where every company that was involved in software, especially if you're a consumer company, knew that you better figure out what mobile meant to your business. And a different way in which I crystallize. How it is a horizontal is by saying that for pretty much the first time in sort of computing's history, there is now a seamless, fluid way to go between

• 4:35Speaker 01

  different kinds of data, whether it is unstructured text to extracting structured information from it, or going from an image to the structure of images, same with video. Things that honestly that you had to do software engineering projects for, or you had specialty areas in computer science that just went and did one thing. All of those capabilities are now exposed to you via AI models. So to make it super concrete, let's say we have a health provider as a Snowflake customer. And let's say they have a bunch of clinician notes from years and years of doctors talking to patients.

• 5:18Speaker 01

  If you wanted to ask a question like, "Oh, make me a list of all patients that had, I don't know, cough symptoms." That would be like a custom job. You had to go find some software engineers, they would figure out what model to use, they would tune it. And then maybe three, six months later, you got like this other table that said, "Okay, this is the list of patients that had cough symptoms." Now, a seventeen-year-old kid can write a little prompt and then run a SQL query on Snowflake. That'll be like, "Okay, here's your list of patients that had cough symptoms." And so first we said, "We want to bring the power of these language models and make it available to anyone that touches Snowflake." And that is Cortex AI.

• 6:03Speaker 01

  It's the model garden, but it's integrated into our Python environment, into our SQL environment, so that you and I can just like write SQL and like underneath it's AI. You don't need to know about where and how the models are hosted, how fast, how slow. We take care of all of that. It's that baseline of capability. And then beyond that, we think in terms of what are like key primitives that use language models, that use AI models, but provide additional functionality.

• 6:31Speaker 01

  So for example, we implemented search. And this was tech that we had developed at Neva. But it's now part of Snowflake where you can create an index on, say, a bunch of documents. We take care of things like chunking it and, you know, generating embeddings, being able to do fast retrieval, so on and so forth. But that's a key primitive that makes chatbots possible. Similarly, we have another primitive for how do you go from spoken text or written text to getting back structured information. And so our vision very much is that there is going to be this base layer. There are important primitives and people can mix and match them to do all kinds of interesting things. They can build applications, they can do data transformations.

• 7:16Speaker 01

  And that, you know, those are things that are still evolving. Absolutely, you can do agents, but part of what we're trying to figure out is like, what is the boundary of where an agent is still reliable? We place a lot of emphasis, for example, with the chatbot. If you want a business user to use it, you want to be able to tell them, you know, you can engineer this, like it'll be right 99% of the time. And if you make a change, you can verify that it's still right 99% of the time. Same goes for the analyst. And so, but we think that by building very, very reliable primitives, we unlock the power of AI for business applications.

• 7:53Logan Bartlett?

  And so, I guess, as you, as you think about instilling this DNA within the organization, or I guess it's probably pivots too strong of a word, but distilling down a focus on AI, I'm not sure how you phrase it, what the right phrasing is. How do you actually do that from a change management standpoint within the organization?

• 8:11Speaker 01

  You know, when the Neva acquisition happened, we, first of all, you know, made sure that the team and us, me and Vivek, my co-founder, were very clear about sort of our mission. We shut down Neva, the consumer product. And we told our team, our goal is to make Snowflake successful. Okay, that's, that, to me, that part is important. Where everybody knows that you don't have an agenda, you're here to make things better. And then, you know, the founders, you know, Benoit Theorie, Christian, our product lead, were very cognizant of the power of AI.

• 8:55Speaker 01

  I mean, at the end of the day, you know, you don't have to be a rocket scientist to internalize what I just said about AI is like a magic data transformation tool. Like, they get it, and they're like, we need to make sure that it's available. But much of the work that we did, to be honest with you, was a nice and to what the rest of the Snowflake team was doing. It was not like that or this. We very much said, this Cortex AI makes Snowflakes equal better, makes Snowflake better. These operators make it easier for people to use the data that is in Snowflake. And in fact, the thing that I tell our customers is, you know, listen, Snowflake does not compete with OpenAI when it comes to foundation models.

• 9:39Speaker 01

  Let's face it, you're simply nowhere as capable of as them. But on the other hand, if you want to use the power of AI to create a chatbot, to do data transformations, to get access to structured data, we'll make it dirt simple for you. And this message of, take Snowflake with all of its glory and power and enhance it with AI, it's like it's a positive cultural message. It's not us coming and saying, you people don't know what you're doing, and let us show you how. That's the way in which a startup fails. I've had plenty of founders that have come into my teams with that kind of attitude at Google. And so in many ways, all of the dumb stuff that we did at Google were like good lessons for us in terms of how do you approach change culturally?

• 10:23Speaker 01

  You know, organizations adapt. And the Snowflake team very much sees this as a say yes and. And that's a message that goes well.

• 10:35Logan Bartlett?

  Cortex AI, have there been any non-obvious ways that people are using it that have jumped out to you, or anything you would point to?

• 10:46Speaker 01

  People are endlessly inventive. And it's the composability that's also super, super cool. We have run into applications, you know, the obvious. Siemens using us because they make 100,000 devices and searching through PDF manuals of 100,000 devices. Like, you know, as anyone that's looked for a washing machine manual knows, like, kind of annoying, kind of hard. In some sense, that's an obvious sort of use case. But like even internally, people have done stuff like, oh, we can annotate use cases sitting in Salesforce. This is just like, we record every new thing that we want to launch within a customer as a use case.

• 11:30Speaker 01

  And tagging them with like the right product, with the right feature is like, you know, is error prone. And also sometimes our sales teams don't do it. But what used to be difficult before and manual tagging, as I said, it's now like, it's like a single sequel prompt to be able to do this. We have hedge funds that are using it to grind through analyst reports and say like, hey, if this company gets mentioned in this report, we will be able to, we'll be able to extract it. There are lots of different ways in which people are mixing and matching what you can do with, with different kinds of models. We also have great partners like Landing AI.

• 12:12Speaker 01

  And that is sort of truly magical because they run within our security perimeter. But they can do things like analyze images and extract structured information from it. It's that sort of composability that is incredibly cool to see.

• 12:29Logan Bartlett?

  You actually launched Arctic, which is your own large language.

• 12:32Speaker 01

  It's our own, it's our own MIE foundation model. And it feels like we did it like 100 years ago.

• 12:37Logan Bartlett?

  Oh, is that right? When did it actually launch?

• 12:39Speaker 01

  I think it was March of this year.

• 12:40Logan Bartlett?

  Okay. And how much of that was, I guess, a customer pull versus the ability to go in and get some native perspective and increase your bona fides in building your own large language?

• 12:51Speaker 01

  100% the latter, which is, you know.

• 12:57Speaker 01

  Like, I think that actually knowing how to use language models is a, I would say, it's almost a required computing skill for any meaningful, any meaningful like infra-level company. Just like, you know, at like the turn of the century, like dealing with big data, like data that could no longer fit in a machine, like, you know, you kind of needed to know how to do that. We thought of Arctic very much as what will it take to train a modestly sized model? Okay. We have talked about it, we spent a few million dollars on it. Again, nothing like the billions that that is being, that are being spent now, but I

• 13:41Speaker 01

  think it was a very good experience for us. And it also taught us a lot about what does inference mean? What does inference optimization mean? And those skills continue to come in handy. We have folks from teams like VLLM, which pioneered a lot of efficient inferencing work. And so I'm pretty happy that we have expertise up and down the AI stacks.

• 14:04Logan Bartlett?

  You guys sit in an interesting nexus point and support a lot of different models, including Arctic, but Llama and Mistral, and you have a bunch of great partners, Rekha, a bunch of great partners or solutions that you all work with. Is there anything you're seeing that's interesting on the model selection side or how end customers are kind of choosing to work with different models?

• 14:25Speaker 01

  I would roughly say for very, very common functions, I don't know, like sentiment detection, people know that small models work just fine and they're sort of happy to go with defaults, it's very inexpensive. But then I would say that when the requirements are a little bit more stringent.

• 14:49Speaker 01

  People like the name brand mark.

• 14:51Logan Bartlett?

  Interesting.

• 14:52Speaker 01

  Which is sort of, it's a weird setup. I think it's the reason why, for example, OpenAI continues to get like heavy, heavy, heavy usage. Yes, they make the process of working with them easy. But I think like people like to go with something that is known and trusted, even if it's just, you know, like a smidgen beyond what you actually need. But why bother figuring out what you actually need when you can just use the name?

• 15:22Logan Bartlett?

  Yeah, the default. It's actually, it's an interesting point that I've seen for sure. Is Anthropic or whoever it is will launch all these new solutions, models, and it pick up a percentage point here or there. But it actually, people just seem to default. I'm sure you saw it in Search. Human behavior is defaults in some way.

• 15:40Speaker 01

  Default's just like, they are very powerful. And there is, it's surprising, but yeah, there's first mover advantage sort of even here.

• 15:49Logan Bartlett?

  As you sort of look at people trying to get applications or into production or leveraging all these different models, what do you see as the biggest pain point in the stack or the value chain to actually get the value or the potential of a lot of these things?

• 16:04Speaker 01

  I think there is still a big gap when it comes to, here is an engineered solution. I think, and I don't think the kind of AI industry, if there is such a term, does itself any favors by simply not talking about things like hallucination rates. Okay, like if you look, no one publishes hallucination rates on their models or on their solution. It's like, look, you're so cool, you should just use us. But I think once you You know, I was actually showing uh um my board members um an app that we have

• 16:46Speaker 01

  that um we can um uh ask questions about our financial data. It's pretty cool because I can say things like um, hey, how did James Better, who is our JVP of uh EMEA? How has he been doing over the past seven days? And it'll like give you a table, it'll run the sequel query and show you um things like we can't be growth and year on year growth and apps, you know, numbers and so on. But before we give it to my uh CFO, Mike, you know, I'll tell the team, you better be sure that um you're you're doing the right thing. I think it is that engineering mentality um that is still a little bit of uh of of an issue.

• 17:26Speaker 01

  And so the more um part of what we end up doing uh is educating our customers on, okay, so you need a release process. You need to actually um, you know, be sure uh that something like ninety five percent accuracy is okay for, you know, okay for you. Or if you use a chatbot, you need to make sure that you provide the citations or place in the document where you can go verify the answer. Because business use like, you know, people take seriously, you can't make uh, you can't make mistakes. Especially when it comes to critical applications, for example, like the one that queries um revenue data on which people can actually make decisions.

• 18:06Speaker 01

  I think it's stuff like that um that um that, you know, continue uh to be obstacles for broad, broad use. But on the other hand, I think AI AI has a ton of use cases to solve problems um where you will happily accept a certain amount of uh mistakes. So for example, if uh uh, I don't know, somebody's, people send articles all the time, right? Um, I find it annoying, it's sort of virtue signaling, they're like, here's a great, here's a great article you should read. You know, my head hurts, why, why another one? So oftentimes, I'll copy the text in the article, paste it into into Claude or into ChatGPT and be like, you know, can you give me a summary?

• 18:52Speaker 01

  If it's off a little bit. I don't really mind. It's not a huge deal. It's not a huge deal because the just the time saving that I get from not actually having to crank through the article, there's like so much value. Similarly, the example that I gave you of my team, I think they updated 5,000 different use cases just by running a SQL query, figuring out, you know, a bunch of product categories and updating the underlying table. I could never have persuaded them to do this manually. And they're like, off by a little bit, it's fine.

• 19:26Logan Bartlett?

  Do you think for mainstream enterprise or business adoption, we need to get to a rate of, I mean, is it 99%? Is it 100% for for it to actually be trusted across an enterprise?

• 19:37Speaker 01

  It depends on the application. So for example, with something like a finance chatbot, I'd be like 100%.

• 19:45Logan Bartlett?

  Yeah.

• 19:46Speaker 01

  Please be, you know, please be accurate or begin to think about how errors are evident. Remember, the the insidious thing about hallucinations is not that the model is getting 5% of the answers wrong. It's that you don't know which 5% is wrong.

• 20:03Logan Bartlett?

  Yeah.

• 20:04Speaker 01

  And that's like a trust issue. And I think so it's going to depend on the application. But this is the good news for all of us as technologists is that there are just a whole pile of these applications where AI will do stuff that you and I have like zero patience for doing. I love the little, I'm a Gemini subscriber as well, and I love the little button that that I get on my Gmail's that says summarize because for things that I want to read quickly, I will happily take, you know, the the few problems that I that I run into. So I think like there is, there are lots of places where the stakes are not that high,

• 20:49Speaker 01

  where AI is already useful. But I think it's also important that we understand that there is a software engineering aspect to how do you create reliable applications like in quite a few contexts.

• 21:03Logan Bartlett?

  When you when you look off in the future, and I don't know what time horizon you sort of think about with with the artificial intelligence industry, it's obviously. Moving pretty quickly, but is there a point in time you pick and say, at this point, I want Snowflake to be known as XYZ in two years' time or five years' time, when you're sort of setting the vision for what's possible within Snowflake with regard to AI?

• 21:27Speaker 01

  Our aspiration is to be an end-to-end solution for data. We want to make it easy for people to ingest data, to be able to engineer it, because data that is collected at source systems, for example, logs, logs will never, you know, logs will always have IDs. Why? Because there's a lot of them and you don't want to be putting strings into it. And so we want to be there when people sort of engineer data, make it more usable, and then obviously when they store it and want to run analytics on it, or when they want to get insights with a dashboard or an AI tool.

• 22:07Speaker 01

  I think we are well on our way to having AI, first of all, play a role in each and every one of these. How do you make it easier to write a SQL query? To how do you make it easy to set up a pipeline? To how do you help people generate metadata for a table or update metadata for a table? I want to write a description for a table, says no one, like, just don't like doing stuff like that. How do we have tools that make it easy? And then, yes, how do we have business users get access to the data that is in Snowflake? You know, my take is that, just like you and I, don't really think, I'm doing this on my phone, just like, do it, because it's natural. That's my aspiration for AI with Snowflake.

• 22:52Speaker 01

  If you feel like you need to do something, and AI is the way to do it, you don't actually think about it, you just go do it, because it works out of the box.

• 22:59Logan Bartlett?

  So to say it another way, when we stop talking about AI and it's just embedded into the process in which we do.

• 23:05Speaker 01

  And it's just embedded into what we do. To me, like, that is actually huge. There are just a whole set of things, again, like, in my own work life, personal life, where AI is just embedded. Like, I don't think about it. I'll give you like, any number of examples. If I want to generate a meme to make fun of someone, or to just say something silly, I don't think twice, I just go, right, I'm very good with words, I just go like, you know, write out a bunch of words, and out will come beautiful memes that I can just like, text someone. You don't need to think about it anymore. Similarly, like I never ever write sheets or Excel formula anymore. I don't know if you like writing them. Total pain, I always get them wrong. I just write English descriptions.

• 23:50Speaker 01

  And these, these, these, these models, all of them are magical at generating code to be able to do stuff like that well. And so I think it just like blends into the background. Because a whole lot of especially soft programming things that like you and I always do. It's like so much easier.

• 24:09Logan Bartlett?

  You obviously lived through the Google and the Internet transformation that society kind of went through. I'm curious, as you sort of look at the potential of AI as it currently stands and the hype that's around it. Do you think we're at a technological point that the hype and the possibility are commensurate with one another right now? Or do you think that we need to make some substantive process on the models or whatever it is to actually meet where we are from a hype standpoint?

• 24:41Speaker 01

  My first answer to that question is like, I, I've been around enough to know that there's a lot that's unknowable. I ask people like, you know, could you reasonably have predicted Uber from the beginning of the Internet? Or even like, you know, when the BlackBerry was around? It's pretty tough, right? And so I think there's just lots of stuff that is quite unpredictable. But I think that predictable parts of AI have to do with what I said earlier, which is data transformations of different kind, new kinds of interfaces.

• 25:21Speaker 01

  And these are pretty easy and reasonable prediction that every app on our phone, or perhaps the phone itself, is going to have a very different interface, like much more of a voice interface. I mean, think about it, even the act of saying, you know, I want to go to this location and I want this kind of a car. And, and once I get into the car, like I'm going to text my spouse and say, this is when I'm going to show up. It's like, that's the kind of stuff that is simply going to change in a pretty dramatic way, all driven by voice. I think that kind of, you know, data transformation is going to be different, is an easy prediction to make.

• 26:05Speaker 01

  And I don't think it requires sort of rocket science technology. It's application of like existing models, yes, with a lot of sweat to integrate them, put them on the phone, figure out what is running locally versus what is central. So I think those are the easy predictions. On the other hand, will we, you know, evolve into this world of autonomous agents everywhere, doing things on our behalf and like arguing with each other and suing each other on our behalf? I think that's a little bit harder to predict.

• 26:39Speaker 01

  Yeah, I think it's an area that's just evolving so rapidly that some of those things are hard to predict. My favorite one is like, what if you had an AI that was designed to get itself more GPU so that it could train better versions of itself?

• 26:57Logan Bartlett?

  Very referential, self-referential in there, yeah. It's a, it's an interesting one. I'm curious, I guess, a lot of the companies we referenced, be it Uber or Google, we sort of ended up in a fairly consolidated power, or at least like user, from a user standpoint. I'm not sure if the market splits and how it all broke out. Do you think we're headed for more federation in the large language model than what we saw in other technologies? Or you think we'll sort of follow a similar trend?

• 27:28Speaker 01

  You know, earlier this year, like, I was actually hoping that we would end up with dozens of model makers in the world. Okay, as we get to the close of this year, it looks like, huh, three, maybe four.

• 27:45Logan Bartlett?

  Yeah.

• 27:45Speaker 01

  It's kind of sad.

• 27:46Logan Bartlett?

  Yeah.

• 27:47Speaker 01

  I actually do sincerely hope that it, it goes back to dozens of companies. I just think it's healthier.

• 27:56Logan Bartlett?

  Yeah.

• 27:57Speaker 01

  I think whether we like it or not,

• 28:02Speaker 01

  concentration leads to, yes, it leads to great companies. I work for Google. They paid for my house. Like, it's like magical. But on the other hand, you know, I think of Effectively call it Google, Facebook and Amazon as kind of like basically a tax on our attention all around. Like they, they, they sort of, just those three companies, you know, collect tax on how we get to information, how we get to entertainment and how we get to products. You know, like, I just think that more competition is helpful

• 28:45Speaker 01

  in terms of how to make more of the benefits of technology flow more broadly than what can happen when there is such heavy concentration.

• 28:56Logan Bartlett?

  Yeah, there's a, I mean, it sort of comes down to antitrust and consumer harm and all those definitions. Because Google, actually, you benefit in a lot of ways for Google being a better, more people using it makes it a better product, which makes it a better service for me and Uber, same thing, the density of, but tha- that's in the near term. Over the long term, uh, who knows? I think competition is probably a good thing for all those businesses.

• 29:20Speaker 01

  That's correct. That's right. That's right. I mean, anytime there is, um, anytime there is massive concentration,

• 29:28Speaker 01

  I, like, I don't think broad society benefits. A, a small set of people that are part of the company, and let's face it, you and I are sometimes part of them and you're very happy that we are part of them, benefit, but, um, as I said, that's more of ish. Uh, and these things also seem to change every six months. So, you know, another six months, hopefully we'll have a different prediction.

• 29:46Logan Bartlett?

  We're wearing background, yeah. I'm curious, when did you start diving into, to AI in a meaningful way?

• 29:54Speaker 01

  I've been in and out, um, in the sense that, um, I wasn't part of the team that created the first, uh, it was a logistic regression model. Except that the team that built it didn't know that they were building a logistic regression model. They were just like strung together a feedback loop and said, uh, we want to estimate, uh, probability and here is how we can do it. But I got to be part of the team like probably early 2007. And so I've lived through a number of revisions of how that product evolved. So we were doing 15 minute updates to these models back in 07.

• 30:38Speaker 01

  Because we had to deal with actual problems like massive spam attacks. We'd have these bots that would just like slam us with clicks. And so if you didn't react fast enough, like the P C T R models would just go stir crazy. And so you had to do real time updates. Or we would have launch events where a new product comes along, massive interest. What does your problem, you know, P C T R model do? So we learned a lot. We also launched, my team also launched the first generation of deep learning models for logistic regression. This is like again, 2011. You know, but by that time, I was also more involved in the business of advertising and commerce and less on the technology of it.

• 31:22Speaker 01

  But but even by 2015, 16, you could begin to tell that there was magic in the air because that's when models started doing completely strange things like being able to translate from, let's say, like English to German while being trained only on translations from English to French and French to French to German. They were learning something about the structure of languages that even the people that were training the models had no idea. And that's when like, you know, we knew that there was something magical there. But then I left Google, started this search engine, Neeva.

• 32:06Speaker 01

  And a lot of early Neeva was really about just like the grind, the mechanics of how do you crawl the web? How do you actually have an index that works? How do you make index selection work? And I would say, again, roughly in 2021, we knew that there was again, something amazing because that's when G P T two came out. It's not that impressive, but early G P T three, even before Chat G P T, it was clear was like a step change. Something went off in our heads when we realized that, you know, this was the key ingredient.

• 32:44Speaker 01

  For being able to create a truly like AI search engine, as opposed to a search engine that returned that returned links. So I started spending more time, you know, with with language models at Niva. We famously pretty much served

• 33:03Speaker 01

  the search experience on seven billion parameter models. We're like, these are cheap. We can run them on A10s and it's enough. And so we'd fine tune, you know, models that open source models that we could then use to power the power the search engines. And, you know, I've sort of stayed in touch. So I've sort of gone in and out of AI and machine learning for the better part of the last 20 years.

• 33:26Logan Bartlett?

  I'm curious if there's any viewpoint that you've changed in a meaningful way about how AI might play it play out. And so, I mean, we talked about the example of different models. Maybe we'd end up with 12 different models and that being very federated. And it seems like we've centralized around a few. Are there any other things that you, you know, have just been very different than maybe how you would have projected?

• 33:47Speaker 01

  I'll give you concrete examples. I think in

• 33:53Speaker 01

  the spring of 2022, GPT-3 had come out, not four. ChatGPT was still like close to nine months away. You know, we started brainstorming on what could AI do for search. Back at the time, you could reliably say, you know, language models can do a pretty good job of what's called abstractive summarization. Where you take a big honking piece of text and instead of extracting words or extracting a sentence as an answer to a question, they do a job of like a first principles summary of like 1500 words. It's like it's magical. And if you think about it, that's crazy hard.

• 34:39Speaker 01

  And so we knew that we could make things work at the level of a single page.

• 34:45Speaker 01

  And if we, Vivek, others in my team said, but wait.

• 34:52Speaker 01

  Can we summarize like, you know, all 10 of the results that we would want to show for a search page into one answer? And better yet, provide citations for where each sentence came from. When I first heard about this in August, honestly, it just looked like absurdly difficult. We went nowhere, using the best models available at the time. And then ChatGPT came out in December, sort of, you know, like just generated a frenzy of hard work and just like brainstorming even within our team about what is possible. On January 6th, we actually had a,

• 35:34Speaker 01

  you know, launched a search engine that did citations, but was powered by things like real-time search. Something, something could explore it on the internet. Five minutes later, you could query it and you could get an AI answer that had real-time information, that had citations. And honestly, I didn't think this was possible just like five months earlier. Or I think that is, that is, that is absolutely something about how innovation in one place just spurs on a whole bunch of follow-on things by lots of people.

• 36:15Logan Bartlett?

  It's interesting. I want to shift gears a little bit to operating and I guess running the business of Snowflake. So Frank Slootman, legendary CEO. I think anyone that's read his book or heard him speak knows he's a very hard-charging, sales-oriented leader. I've noticed, or I've sensed that since you've come in, there's been a real focus on product velocity and shipping in a more meaningful way than maybe there was in the past. My words, not yours, that, that just focusing on incremental improvements or even launching products in a more meaningful way and just being willing to try things and fail.

• 36:57Logan Bartlett?

  I'm curious, how do you set the culture of that? And how do you think about taking the risk of launching new things that might not work out?

• 37:08Speaker 01

  I mean, you asked, you know, I would actually say that.

• 37:15Speaker 01

  You know, my overall message to the Snowflake team is that this is a time of unprecedented change in data as a whole. First, data is sexy, which means that there's lots of competition always, whenever stuff like that, stuff like this happens, and there is a ton of technological change as well. My overall message to the Snowflake team is you got to step it up every, you know, in every department.

• 37:43Speaker 01

  And I spend a lot of time with sales, I spend a lot of time with customers. And so, you know, in fact, I would say I am driving a process of being way more quantitative and organized about how the sales team operates. And, and to me, one of the most difficult skills that only some people end up mastering that I'm proud of, you know, for having developed is how do you do like scaled process management, which is sort of what, which is what, which is what sales is. How do you bring the power of computing, the tools that, that you have, to just be a more effective team?

• 38:25Speaker 01

  And so I would say that sort of, hey, we need to step it up because this is a defining moment, is throughout Snowflake. Absolutely. On the product side, I think I want to make sure that we have, that we have clear focus. Snowflake came of age at a time when they were like five, six years ahead of everybody else. It was a magical product, absolutely sold itself. But there's a lot more competition now, for good reason, that is a big market. And so this is part of the reason why I stress that sense of urgency. I also stress very much the need for virtuosity in everything that we do. What I mean by that is like, look, there's only so many things that are knowable in

• 39:10Speaker 01

  the world ahead of time. And the things that great companies, you know, I was part of Google, which is one of the greatest companies ever, is not that we had all the answers, but we were way faster than anyone else in like going and seizing that sort of opportunity. And the big difference between like, you know, cloud computing and shipping, you know, CDs was could iterate every day. You could do multiple releases every day. And so that just provided. And the entire company, by the way, running a $50 billion business, $100 billion business,

• 39:51Speaker 01

  would just go update all of its binaries every week from head. Just like the pace of change was truly incredible. That's again, part of what I tell my team here, which is you will not know all the answers and that's okay. And you got to figure out how to iterate. That doesn't mean that we iterate at like the level of storage formats. If you're running a database company, like, don't mess with my data is rule number one. You have to preserve that. Like, don't let hackers break in. Also, rule number one. So there are some inviolables about how you operate, but for a whole lot of other things, you got to be really, really nimble at getting things out because you just get more feedback. You're able to, you're able to see how customers use your product.

• 40:36Speaker 01

  You can get feedback from them. You can iterate a whole lot faster. Like, honestly, it's a lot more fun for the teams as well.

• 40:45Logan Bartlett?

  I've heard from an accountability standpoint that you, you ask or force teams to make promises on deliverables for, is it two weeks and three months? Or how, how, how does that typically work?

• 40:56Speaker 01

  They get to pick.

• 40:57Logan Bartlett?

  They get to pick the time, right?

• 40:58Speaker 01

  You get to pick, but it's, I like, I do not want to listen to any, you know, FY26 promises. Kind of like, yeah, I don't, I don't do that. I'm kind of like, you want to, you want to sign up for what you will get done in a quarter? It's fine. Please do a good job. I like truly stated goals. I argue a lot about goals, but clearly stated goals. I like responsible individuals. I ask like, is it one person? Is it two people? But your name is against that goal and your promised date is also against that goal. And then, you know, we touch base every two weeks just to check in on things. How are things going? You know, what needs to be tweaked, so on and so on and so forth.

• 41:41Speaker 01

  But the first three, the goals, the owners and the dates don't change. If you're late, like, yeah, it's awkward, but it's okay. Deal with it.

• 41:50Logan Bartlett?

  I've heard you've also, you have a 5:00 AM email digest that comes through. Is that, is that something?

• 41:56Speaker 01

  That's for our revenue, absolutely, every morning.

• 41:59Logan Bartlett?

  And is that something you've done in the past?

• 42:02Speaker 01

  I've done that in the past. As a joke, a friend of mine and I set up a mailing list at Google called Bean Counters. We're like, you know, we count the beans. That would come every morning and it would say how each product category did. It became a little bit of a ritual. It got adorned over time with like spark lines for how things were trending and so on. But we instantly knew if there was a problem in an area or if some area was doing well. And I instituted that here as well. So we get a report of how the company did. Every day, we're a consumption company. We make money when our customers spend money.

• 42:47Speaker 01

  And so, yeah, that's, that's sort of, it's a tradition and a bunch of myself leaders, you know, get it. It's a little bit of accountability.

• 42:56Logan Bartlett?

  Yeah.

• 42:56Speaker 01

  And things go wrong. I want people to step up and go dig into it and find out why things are going wrong. Or if things, you know, get better as well. I think it's just, it's, it's sets a tone of, you know, we, we are a business. And paying attention to the things that matter is just what we do.

• 43:17Logan Bartlett?

  I'm curious, you talked about like that as an example of something you did dating back to Google and Bean Counters. I'm curious, are there notable ways that your leadership style, you feel like, has changed since Google or with Neva or now within Snowflake? Anything that stands out that you wish you could have told yourself way back when?

• 43:34Speaker 01

  Yeah, I would say one big. Big, big change from, I guess, my early, like, I was a reluctant leader. I used to tell people at Google that, you know, I became a manager or a director so that I didn't have to listen to people tell me what to do. It was like my own ego. And, you know, over time, you realize, no, it's actually a responsibility, you're privileged to, you know, to, to, to have a team and be able to steer it and get things done that you can never do as an, as an individual. But I would say the early me was, you know, quick to criticize, quick to set up a lot

• 44:14Speaker 01

  of like win-lose kind of scenarios, which in my mind just are not, you know, not, not that helpful. I would say I'm just a whole lot better at not creating unnecessary conflict. At, if I have a point of view or if I think that something is not going right, being able to surface those in ways where the other person realizes that things can be done better without my having to like dictate to them that things are wrong and I am the one that has the, that has the solution. I think I'm getting much better at, at like, you know, getting people to get to the right answer rather than me being the person that says that either something is wrong or that

• 44:59Speaker 01

  I have the answer. I have to credit people like Jeff Huber, who was my boss, gosh, for close to 10 years at Google. He always was a big fan of like the, the, the Socratic dialogue kind of technique to get people to reach the conclusion that they should be reaching.

• 45:18Logan Bartlett?

  It's interesting. I'm curious, you've seen hyper scale and hyper growth a few different times in your career. If, if someone's listening and they're earlier in their, their journey but have the aspirations to be the next Google or snowflake.

• 45:34Logan Bartlett?

  Are there things that you would encourage people to think about today to set the right scaffolding or foundation in the early days so that you're not forced to course correct on down the line?

• 45:45Speaker 01

  I think course correction is okay. I think, you know, some things are fatal, I agree. I think like a positive blameless culture, a culture in which the benefit of doubt is always given, a culture in which you expect people to do the right thing because that's the right thing to do. I think those things are foundation. You don't want a blame culture. You don't want, you know, you don't want a suspicious culture, like, you know, like they're not going to do the right thing. I think those things are a baseline.

• 46:25Speaker 01

  But beyond that, I think having both a sense of paranoia that success can be fleeting or it can be taken away pretty easily. And, you know, having a sense of are you really doing the best that you can be? Are you being excellent in how you approach stuff? That ability to self-correct on a continuous basis, I think is important. I think it's also like, you know, you have to accept that, especially in startups, almost by definition, you're doing new things. I think you'll make mistakes. I think it's okay to make mistakes. I place a lot of value in like, accept your mistakes, don't make them again. I used to have a little, you know, like one of these magnets in my cube at Google that

• 47:10Speaker 01

  would say like, make new mistakes. And I think having that culture of you will accept failure, you will make mistakes, you will correct, you will learn, and you will work towards a different, better future, is what I would tell people.

• 47:24Logan Bartlett?

  One last one before I let you hop. I've heard you mention that people shouldn't necessarily follow their passions, but should instead find the intersection of their strengths and what value they can potentially bring to the world. How do you think about if someone's Mapping out what their career could look like over the course of, you know, the next decade or whatever it is, how do you think about picking a career and pursuits?

• 47:52Speaker 01

  I'm not expert, but I'll tell you, I'm also painfully aware that we live in a society that sort of idolizes very unique individuals. Okay, like, let's face it, Steph Curry is our hero.

• 48:04Logan Bartlett?

  Yeah.

• 48:06Speaker 01

  Or, I don't know, Instagram or YouTube influencers with a massive follow, Mr. Beast. And like, my main take is that, you know, people's passion often points to these pyramidal structures where like, you have to be extraordinary to be, to be successful. I'll, it's an outdated stat, but it will illustrate the point.

• 48:35Speaker 01

  You know, a few years ago, if you counted the number of, the number of creators on YouTube that made like 50K or more a year, as you know, it's hard to make videos. It's hard to get a lot of followers. And $50,000 isn't a lot of money if you want to support yourself or a family. The number of creators on YouTube that would make that kind of money wasn't the tens of thousands. Okay, it's not the five million lawyers we have, or the six million software engineers. It's not a scaled occupation. And so to me, it is really important that people steer towards, like, they, they sort

• 49:13Speaker 01

  of play the game of life in a way that guarantees at least modest success, than pursue these elite professions in which like, let's face it, there's one Steph Curry. It's really hard to be him, I mean, good for him, but it's really hard to be him. That's sort of more of what I mean by, you gotta be smart about how do you steer your life to a direction that combines your interests with what society also finds broadly useful. Otherwise, you end up with like, really difficult challenges.

• 49:48Logan Bartlett?

  Well, thank you for doing this. This is great.

• 49:50Speaker 01

  Thank you.

• 49:53Speaker 00

  Thank you for joining this episode of the Logan Bartlett Show with CEO of Snowflake, Sridhar Ramaswamy. If you enjoyed this discussion, I'd really appreciate it if you subscribed on whatever podcast platform you're listening to us on, as well as share with anyone else that you think might find it interesting. We look forward to seeing you back here with another great guest on the Logan Bartlett Show next week. Have a great weekend, everyone.