Snowflake Summit 2025 | Russell Spitzer, Snowflake and Apache Polaris & Laurent Bride, Komodo Health

Clips
More from Snowflake Summit 2025

Russell Spitzer

Principal Software Engineer

Snowflake & Apache Polaris

Laurent Bride

Chief Technology Officer

Komodo Health

play_circle_outline Komodo Health's mission to reduce burden of disease by leveraging healthcare data and AI.

play_circle_outline Governance strategy around open table formats and use of Apache Polaris for security.

play_circle_outline Collaboration between Snowflake and Komodo Health on data processing and analytics.

play_circle_outline Challenges and future enhancements in the Iceberg community for smaller tables and faster writes.

Info
Transcript

Russell Spitzer, Snowflake and Apache Polaris & Laurent Bride, Komodo Health

Russell Spitzer

Principal Software Engineer Snowflake & Apache Polaris

Laurent Bride

Chief Technology Officer Komodo Health

In this Snowflake Summit 2025 conversation, Laurent Bride, CTO of Komodo Health, and Russell Spitzer, software engineer at Snowflake, join theCUBE’s Dave Vellante and John Furrier to unpack how open standards are reshaping healthcare analytics on the AI Data Cloud. Bride explains how Komodo Health’s full-stack analytics platform pairs the industry’s most complete view of de-identified patient journeys with Snowflake’s performance engine, activating Apache Iceberg tables to democratize insights and reduce disease burden at scale.

The discussion dives ... Read more

explore Keep Exploring

What is the mission of Komodo Health and how do they work towards reducing the burden of disease in the healthcare ecosystem? add

What is the governance strategy around governing an open table in regards to new technologies such as Iceberg and Apache Polaris? add

What kind of relationship do you have with Snowflake and how do you communicate with them? add

What are some proposed improvements for Iceberg V4 focused on smaller tables and faster writes in order to compete with other formats like Delta and Snowflake? add

bolt Powered by CUBE AI

Russell Spitzer, Snowflake and Apache Polaris & Laurent Bride, Komodo Health

search

Rebecca Knight

>> Hello, everyone. And welcome back to day two of theCUBE's Live coverage of the Snowflake Summit here in San Francisco, California. I'm your host, Rebecca Knight, alongside my co-host and analyst, Dave Vellante. We've got two great guests for this next segment. We've got Russell Spitzer, welcoming him back, principal software engineer, Snowflake and Apache Polaris at Snowflake. Welcome returning to the show, Russell, for triumphant return.

Russell Spitzer

>> Thank you.

Rebecca Knight

>> And Laurent Bride, he is the Chief Technology Officer at Komodo Health. Welcome, Laurent.

Laurent Bride

>> Thanks for having us.

Rebecca Knight

>> So for our viewers who may be unfamiliar, I'd love you to tell us a little bit more about Komodo Health and what you're all about.

Laurent Bride

>> Yeah, so Komodo Health mission is a very simple but ambitious and humbling, is to reduce the burden of disease. And we do that by providing unique insights into the entire healthcare ecosystem. So we work with life science company like any big pharma company, payer provider. We work with patients, advocacy groups, but we also have startups that build their entire business on top of the Komodo platform. So the Komodo platform is all fueled by the healthcare map. So I'll talk a little bit about that because it's all about data. And then the past two years we've been injecting a lot of AI into the platform. So we really provide novel solutions across multiple market segments and serve many, many personnels.

Dave Vellante

>> So we definitely want to get into your motivation for using open table formats, how are you using them, where Snowflake fits. But Russell, for you, I want to start with the big picture. So last year and a very small sample, we had a quick survey we did of about a hundred IT decision makers just to find out where they are with data lake and open table formats they're using today and in the future. And number one was Hudi today, followed by Delta and then Hive. Actually no, Iceberg was bigger than Hive. No surprise there. But for the future, six to 12 months, Iceberg was number one, about 40% of the customers, with kind of Delta declining, Hive declining, Hudi of course declining. Other actually, I don't know what other was, was maybe Parquet, but so we didn't get into that. But anyway, how has that changed? It feels like the market has really spiked since that survey which we took last I think August.

Russell Spitzer

>> I mean it's tremendous. You can really feel it in the Iceberg community as well. You probably know we had our first Apache Iceberg summit.

Dave Vellante

>> Yes, of course. Well, one of our guys was there, George.

Russell Spitzer

>> It was amazing. I mean, you can't imagine having a sold out conference for a table format. Amazing the amount of momentum that's coming. And I think it's because every engine has kind of jumped onboard and said Iceberg is the important thing to do. All the major vendors are onboard. So if you do Iceberg, it works everywhere. So that's the future.

Dave Vellante

>> Funny. We had Sanjeev Mohan on this morning. He's like, "We're spending all this time talking about Iceberg. We're moving up the stack." But it's an important topic. So Laurent, my question is what motivated you guys to go to open table formats? What was the catalyst there?

Laurent Bride

>> Yeah. So we actually start talking about Iceberg a long time ago to Snowflake because for us, what's really important is to avoid any kind of vendor lock-in. And looking at Iceberg, it really gives us that kind of interoperability. And we love Snowflake. We are using it across the board, but there might be some specific workloads that where we want to use another type of engine. And when I see Iceberg, like you talked about the committee, Russell, like Flink, Spark, Dremio, of course, Snowflake, all of those have adopted Iceberg. So for us, it's the interoperability and the flexibility it offers and to avoid that kind of lock-in.

Dave Vellante

>> Okay, so you can choose the engine to apply to the data, but it's only for... Am I correct that it's for a subset of your data? There's other data you're happy to leave in other table formats. Or are you moving wholesale to Iceberg?

Laurent Bride

>> If you think about the healthcare map, we're talking about petabytes of data that we are processing every month. And we have different version of our healthcare map that's called Fabrics. And here we're talking about terabyte of data. And we actually want to formalize everything and bring everything to Iceberg. So for all of our use cases, ultimately there's going to be an Iceberg format moving forward.

Dave Vellante

>> So the million-dollar question is always, okay, how are you governing that open table? Because you go online, you look in social media, you hear all these little nuances about Iceberg. It's not maybe as trivial as everybody thinks, but that's how new technologies are. So what's your governance strategy around this?

Laurent Bride

>> So catalogs, we've applied a lot of internal rules and policies in place around all the tables we have. But we've been adopting Apache Polaris, which I also appreciate Snowflake giving that to the Apache community I think it was last year or something like that.

Russell Spitzer

>> Yeah. Almost exactly a year ago. Yeah, right?

Laurent Bride

>> So yeah.

Dave Vellante

>> Talked about that on theCUBE. Great discussion.

Laurent Bride

>> Everything security-based, now we are pushing that toward catalogs. Yes.

Dave Vellante

>> So Russell, help us understand. I remember talking to Benoit at one of our super cloud events. I said, "Yeah, but Polaris is really just the technical metadata." And he said, "No, no, no, that's not true. And over time we'll continue to enhance it." So help us squint through all the noise in my brain. It feels like you're moving more of the wonderful proprietary function from Horizon into the open community. Is that a correct understanding? Update us on that.

Russell Spitzer

>> I think so. I think basically what we're seeing is there's a bunch of primitives and a bunch of functions that a catalog really needs to work as a lake house catalog. And that means that there's some functionality that used to just be in Horizon that needs to also be in Polaris. Because for example, if I want to do row-based access control, that has to be in the catalog that every single engine knows how to talk to and every engine knows how to deal with. So we've basically been looking at those things and we decide which ones are important that everyone has the same view of. Obviously if there's something that only exists within Snowflake, not really important. But things like row-based access control, all that kind of stuff, everyone needs it.

Dave Vellante

>> And that's something that he responded to, Benoit, and I think it's because he knows what the development team's working on. He's like, "No, no, no, we're going to be moving in that direction."

Russell Spitzer

>> Oh, yes.

Dave Vellante

>> Irrespective of the actual announcement timeline. So was that announced here? Was that just sort of part of the committer activity? How did that all come about?

Russell Spitzer

>> I mean, I think it's been a day-to-day thing from the moment I joined Snowflake. I had a similar experience where I love Snowflake the engine and I love Iceberg the format. And when I was given the opportunity to come on board and make those things better together and take things that were in Snowflake and make Iceberg better with them, I was all about it. So I don't know if we've talked about it enough out loud, but our goals here are changing the way Iceberg interacts with catalogs and the way that Iceberg exists for everybody. We want to make it just the best solution possible.

Rebecca Knight

>> Well, speaking of changing the world, I want to go back to the mission of Komodo Health, which is an ambitious mission to reduce the global burden of disease through this healthcare map. So can you tell our viewers a little bit about the kinds of data and tooling that you're using and how you're using it? Because I feel as though there's so much inspiring stuff going on in terms of the power of AI to make sense of unstructured data from researchers and life care sciences companies. So just help bring this to life for our viewers.

Laurent Bride

>> So let me take you back to the data in the healthcare. It's really messy. It's siloed, it resides in different system, EHR, EMR, claims, you name it. It's inconsistent, it's fragmented, and often it's of poor quality. So when you look at all of that, you need a way to put that and stitch that all together. And that's what Komodo did. And we started building that healthcare map, which is all about creating that longitudinal view of patient's life. I'm sure you guys see the doctor, I'm sure you are in our database. I don't know who you are, of course. It's all tokenized. But we can replay the entire life of a patient in the healthcare map. And we've used Snowflake to do that. So as I mentioned before, we're talking about petabytes of data that we are processing all the time. And because the data is messy, there's a strong ingestion phase, then there is a reconciliation, a lot of transformation, and we've been using Snowflake to really help us scale on that front and bring the right level security and privacy when it comes to the data we handle.

Dave Vellante

>> How is the decision made? You mentioned Trino and Spark, Snowflake. How is the decision made to bring an engine to the data? Is that a situation of maybe there's a person, a technical person says, no, we're using Snowflake for this or Spark for that? Or is it the parameters or nature of the data set? How is that decision made?

Laurent Bride

>> I think it's a mix of different factors. We started with Spark, we had a lot of Spark engineers, we did a lot of Spark coding. Native Spark coding, it's complicated. What Snowflake really bring is that ease of use and you have way more SQL coders than Spark Java native coder for example. So you can start creating those transformation, cleaning the data, reconciliating all of those records, it was much easier to do that in SQL. So there's the ecosystem of developers that you have. Then there's the cost part. You're looking at managing your Spark clusters and the infrastructure behind that. So there's pros and cons in all of that, but then you evaluate different solutions and you make a decision.

Rebecca Knight

>> Excellent, excellent. And so how are your teams working together? Obviously you are both collaborating on how you work. So tell us a little bit about how you work together.

Russell Spitzer

>> As an open source person, almost all of my communication is with all of the users of Polaris and Iceberg dev mailing list. And anyone out there who wants to get involved, you're invited as well. So we take requests, features, anything you want and we're there to answer and get involved.

Laurent Bride

>> Yeah, I have the direct line cell phone of Benoit and Jerry and that's how we do it. No, I'm just kidding. We talk about our challenges, we talk about what's happening in the market. We mentioned Iceberg before. We said, Hey, we need this kind of format. We are part of that customer panel. So we engage on roadmap on QBRs and then we can influence roadmap and we really see the Snowflake relationship as a partnership and not like a vendor-customer type of relationship. So yeah, we dig into the details and we have a very responsive team and we talk about cost, we talk about performance, we talk about governance, we talk about what's happening in the market, like the different partners we want to work with. And we're bringing the innovation together in the market.

Dave Vellante

>> Russell, when you think about the number of announcements that were made, the fire hose of announcements that was made this week, what aspects of those should we be focused on? Help us extract a signal that is specifically relevant to open source, open table formats, Iceberg, Polaris.

Russell Spitzer

>> I mean the thing that's most exciting to me is that while we were announcing all these capabilities inside of Snowflake, almost all of those that we talked about, Snowflake also had a direct hand in getting into the Iceberg project in the first place. So it's this whole all-in approach. Snowflake believes in Iceberg and is making sure that things go all the way through into the spec. So that's the thing to look at is what are your favorite things in Snowflake that you think everyone should have access to? And that's kind of what's going to come next.

Dave Vellante

>> Okay. So whole house kind of leaning in.

Russell Spitzer

>> That's how I feel.

Rebecca Knight

>> So I want to talk use cases with you, Laurent, because this is a really powerful foundation of technology that really promises or has the potential to make healthcare more accessible and actionable and really improve patients' lives, particularly patients who are living with the burden of disease. So paint a best-case scenario for us here about what you can actually do with this data and with bringing together researchers and companies who are working on these problems around the world.

Laurent Bride

>> Yeah. So maybe I can take an example around clinical dev or clinical trial. So you want to start a new drug and you want to basically find patients, you want to recruit patients to jump on your clinical trial. So right now, you would have a to buy data, like hire consultants, put everything together in analytics. We have all of that in the healthcare map. So here you would just use our tooling. We have a code builder, we injected some AI. You can ask natural language questions and then get your quote of patients that you can start to engage with through a healthcare provider. I can give you another example in the commercial world of life science. You just launched a drug and you want to know how well that drug is going on the market and you're going to do that through evaluating prescriptions. So you would again look at the healthcare map and look at the claims and we have built a set of analytics that will be able to tell you how well your drug is doing. Are you getting market shares around your competition? Should you move to another stage? Should you change your marketing campaigns? All of those use cases we can address today with the healthcare map. And the AI part made that so much simpler. Yeah.

Dave Vellante

>> What's happening in the community, Russell, to bring Delta and Iceberg together? I remember we had Ryan Blue on, George Gilbert and I, well before the Databricks acquisition and he was skeptical. He was like, "That's never going to happen." And of course once a company pays a couple billion dollars for you, maybe that changes your mind. But is there activity in the community and GitHub to make that all go down? I mean I know that's probably not your priority.

Russell Spitzer

>> Yeah, I mean I'd probably say in general the community is not super concerned about that except for certain members who are incentivized to make that happen. But as a community, we accept anyone who wants to be a part. So if someone comes in and they have a good idea, we're willing to accept it.

Dave Vellante

>> Its open source. That's what it's about. Okay. But that's not something that is gaining a lot of momentum, is that what you're saying right now?

Russell Spitzer

>> I mean, I wouldn't think that it's an objective of really the community in general.

Dave Vellante

>> Yeah, and that's what we inferred when we talked to Ryan. And of course post-acquisition, it's hard to squint through. And that's been our assessment, focus on Iceberg, making it simpler. Well, what are the key challenges with Iceberg that the community is working on that we should be aware of?

Russell Spitzer

>> Yeah, I think we have a lot of proposals for V4 and they involve a lot of contributions, some that would make the format a little more like Delta, some that make it a little more like Snowflake. But in general, we're looking at trying to solve problems around smaller tables, faster writes to Iceberg, trying to alleviate some of the things we see other formats doing better than Iceberg. And we look at that and we say we should probably be doing that too.

Dave Vellante

>> So smaller tables and faster writes probably go hand-in-hand, yeah?

Russell Spitzer

>> I'm thinking a lot about streaming and fax tables. So we have a lot of use cases where people have rather small tables but with rapid updates. And those don't work great in Iceberg right now because Iceberg's really a batch solution, works much better with medium-sized tables and less frequent writes and lots of frequent reads. But we're interested in handling that other case as well, so we're trying to figure out what we can do to the format to make it better.

Dave Vellante

>> So Laurent, what do you want from the community, the Iceberg community?

Laurent Bride

>> Like on Iceberg specifically? I think all of it.

Dave Vellante

>> Hurry up.

Laurent Bride

>> Yeah, exactly. I mean for us, there are critical aspects of Iceberg that we really love. When you think about analytics, you want to be able to rerun analytics at certain point of time. So the ability to go back in time, the time travel, if I may, the ability to work on incrementals, the ability to append new data without having to rewrite tables, all of that is going to help us on the cost side, on the efficiency. So whatever you do, make sure that the analytics are faster, that they are costing less from a compute cycle perspective and we'll be happy.

Russell Spitzer

>> Of course.

Dave Vellante

>> Interesting to hear you say that because these in a mature format takes time. I get it. But you have decided that that risk of lock-in, the ability to bring any engine to that data is worth the wait for that. So you must have done a sort of analysis to say, hey, if it took 50 years, you wouldn't do it, so it's got to be sort of near-term, mid-term, long-term. What went into that decision? I'm interested in getting inside your head and when you said, okay, we have enough confidence that the community and the open source is going to catch up to where we need it to be functionally that we're willing to make that bet.

Laurent Bride

>> Yeah. I come from the open source world. I was CTO at Talend for six years, so I believe in the community. I have a lot of friends who are Apache committer. One good thing about Apache, we can jump on the wagon. If it doesn't go fast enough, now I know Russell, I can ask him, hey, can you push that faster? I think we're going to get there. And it's not a big bang approach. We're not moving everything overnight to Iceberg. We still have a lot of data that's in the Snowflake proprietary format and we have other formats. So I don't think we are risking the company on a format, but we know that at some point we want to consolidate because that's what makes sense down the road.

Dave Vellante

>> It's that open source ethos. You drank the Kool-Aid long ago.

Laurent Bride

>> Yes.

Dave Vellante

>> And in this world, that makes a lot of sense. I'm not sure how well it applies to open source chips and CUDA. We'll see. But this world, you can move much faster than you can in many others.

Russell Spitzer

>> Yeah, I mean there's no easier way to get involved than open source. You can just get your hands in whenever you see something you don't like and you can change it. I mean, it's great.

Rebecca Knight

>> Nothing better. Excellent. Well, Russell and Laurent, thank you both so much for coming on theCUBE. A really interesting conversation.

Russell Spitzer

>> Thank you.

Dave Vellante

>> Thanks, guys.

Laurent Bride

>> Thank you very much

Rebecca Knight

>> Best of luck.

Laurent Bride

>> Thanks.

Russell Spitzer

>> Thanks.

Rebecca Knight

>> I'm Rebecca Knight for Dave Vellante. Stay tuned for more of theCUBE's live coverage of the Snowflake Summit. We'll be taking a quick break here and then we'll be right back this afternoon. Come back for more. You're watching theCUBE, the leader in enterprise tech news and analysis.