Panel: Shift Left Across the Data Lifecycle—Data Contracts, Transformations, Observability, and Catalogs | Prukalpa Sankar, Tristan Handy, Barr Moses, Chad Sanderson | Shift Left Data Conference 2025

Talk Description

Join industry-leading CEOs Chad (Data Contracts), Tristan (Data Transformations), Barr (Data Observability), and Prukalpa (Data Catalogs) who are pioneering new approaches to operationalizing data by “Shifting Left.” This engaging panel will explore how embedding rigorous data management practices early in the data lifecycle reduces issues downstream, enhances data reliability, and empowers software engineers with clear visibility into data expectations. Attendees will gain insights into how data contracts define accountability, how effective transformations ensure data usability at scale, how proactive how proactive data and AI observability drives continuous confidence in data quality, and how catalogs enable data discoverability, accelerating innovation and trust across organizations.

Additional Shift Left Data Conference Talks

Shifting Left with Data DevOps (recording link)

• Chad Sanderson - Co-Founder & CEO - Gable.ai

Shifting From Reactive to Proactive at Glassdoor (recording link)

• Zakariah Siyaji - Engineering Manager - Glassdoor

Data Contracts in the Real World, the Adevinta Spain Implementation (recording link)

• Sergio Couto Catoira - Senior Data Engineer - Adevinta Spain

Panel: State of the Data And AI Market (recording link)

• Apoorva Pandhi - Managing Director - Zetta Venture Partners
• Matt Turck - Managing Director - FirstMark
• Chris Riccomini - General Partner - Materialized View Capital
• Chad Sanderson (Moderator)

Wayfair’s Multi-year Data Mesh Journey (recording link)

• Nachiket Mehta - Former Head of Data and Analytics Eng - Wayfair
• Piyush Tiwari - Senior Manager of Engineering - Wayfair

Automating Data Quality via Shift Left for Real-Time Web Data Feeds at Industrial Scale (recording link)

• Sarah McKenna - CEO - Sequentum

Panel: Shift Left Across the Data Lifecycle—Data Contracts, Transformations, Observability, and Catalogs (recording link)

• Barr Moses - Co-Founder & CEO - Monte Carlo
• Tristan Handy - CEO & Founder - dbt Labs
• Prukalpa Sankar - Co-Founder & CEO - Atlan
• Chad Sanderson (Moderator)

Shift Left with Apache Iceberg Data Products to Power AI (recording link)

• Andrew Madson - Founder - Insights x Design

The Rise of the Data-Conscious Software Engineer: Bridging the Data-Software Gap (recording link)

• Mark Freeman - Tech Lead - Gable.ai

Building a Scalable Data Foundation in Health Tech (recording link)

• Anna Swigart - Director, Data Engineering - Helix

Shifting Left in Banking: Enhancing Machine Learning Models through Proactive Data Quality (recording link)

• Abhi Ghosh - Head of Data Observability - Capital One

Panel: How AI Is Shifting Data Infrastructure Left (recording link)

• Joe Reis - Founder - Nerd Herd Education (Co-author of Fundamental of Data Engineering)
• Vin Vashishta - CEO - V Squared AI (Author of From Data to Profit)
• Carly Taylor - Field CTO, Gaming - Databricks
• Chad Sanderson (Moderator)

Transcript

*Note: Video transcribed via AI voice to text; There may be inconsistencies.

" We're gonna bring on our panelists for this next session if they're around. I already see Chad is back.

Good to have you here again. Who else do we have around? Oh yeah. Look at this Tristan bar. Oh, right, everybody's on. That's my cue. I'm out. I'll see you all in a bit.

Alright, um, well, hey folks, I'm back again except this time. I've got some very important people on the panel with me. So without further ado, um, maybe we can just get started with some introductions and I'll let everyone sort of quickly introduce themselves and we can get onto the, to the cool stuff. So, uh, bar maybe do you wanna start with you?

Sure. I love Demetrius Energy gets me going. Um, I'm, uh, I am Bar Moses. I'm the CEO and co-founder of Monte Carlo, uh, the data and AI Observability Company. Um, so what we do is we help, um, amazing companies like, uh, Roche, NASDAQ, Honeywell, many others, uh, deliver reliable data and AI products by providing observability, um, to make sure that, uh, they are aware, um, of data issues can detect them early and resolve them quickly.

Excited to be here today. Awesome, Kristen. Hey everybody. Uh, founder and CEO of DBT labs. Uh, DBT is at this point, uh, the, the default way that data transformation is done in the cloud. Alright, and, uh, last but not least, uh, ALPA. Hi everyone. I am the founder of Alan. We think of ourselves as the home for data teams, so a collaboration control plane where we bring together trusting context across the entire data estate.

Um, we started off with our first use case being sort of like this concept of third generation data catalogs, but what's become super interesting is that started becoming the window for this broader concept of how do you help data teams and data producers collaborate super well with the business and build this broader collaboration plan on everything around data.

Data. Awesome. Well obviously some heavy hitters here, so I think there's a lot of cool conversation that we can have. Um, so, you know, topic of the day is, is shift left and shift left is all about moving from sort of downstream systems, uh, sort of people who are not in the code, not writing code, um, to sort of the origin point of where issues, uh, where these issues come from, which ultimately, um, usually lies, uh, within the domain of the engineering teams.

And, and I think all three of you, when you are building your companies. Had that kind of engineering first mentality, uh, in mind. Um, I'd love to know, as you know, over the past like five years, uh, are you starting to lean more into that? What are you kind of seeing from the market, um, in, in terms of, you know, this sort of engineering based persona applying, uh, data management and, and data workflow, um, sort of tooling for them?

Maybe Tristan, uh, we, we can start with you since of course this was your thesis from day one. Yeah. I've been beating this drum for a long time. Um, the, the original kind of DBT manifesto came out in 2016. And, um, the, the, the problem that I saw then is, you know, there, there's a lot of data engineers that build data pipelines, but there's also a lot of other people that build data pipelines.

And, uh, typically those folks have, uh, quantitative backgrounds, but they don't have engineering backgrounds. And as a result, they created a bunch of messy tech debt that just didn't scale, and it wasn't clear to them, uh, back in 2016 how to do things in a different way because it's just sort like not their background.

And so, uh, the thing that DBT, uh, has, has done over the years is tried to, uh, train a much larger cohort of human beings on how to apply software engineering best practices to, to data. Um, and you know, that's everything from, uh, you know, everything should be in code and human readable code, and it should go through an automated testing process and a pull request process.

And, um, and I, uh, the, the, the feedback that I got in the early days was that nobody really wants this. Like, uh, this is kind of naggy, um, data analysts just need to get their jobs done. And there's, uh, there's no such thing as a, a data analyst who wants to think long term about the work that they're doing as, as systems.

Um, but that turned out to not be true. And, uh, a lot of folks of, of many different stripes in data, uh, do, do want, uh, to, to think about the work that they're doing as, uh, long-lived production software systems and, uh, are very happy to, uh, to have the upleveling of their career that it means to go from somebody who just kind of creates one-off assets to, to somebody who's building production systems.

That's awesome. That's great. Uh, BARR may maybe you next. I I think what I'm, what I'm really interested in the observability space is that observability is, you know, sort of synonymous with, with engineering and, uh, what, what I saw sort of you guys trying to accomplish is like, okay, can we, can we bring the same sort of.

Rigor to quality and monitoring and an and anomaly detection that engineers have into the sort of data engineering, data development lifecycle. Um, how, how, how has that kind of been been going and where, what, what do you think the sort of future is in terms of shifting potentially even further to the left?

Yeah, great question. I, I, um, you know, reflecting on the, how the, you know, the industry has changed in the last couple years, I sort of just echoing, you know, what, what Tristan has said in terms of, um, it seems like the lines between data people or engineers are sort of blurring. And I think AI actually is making that even even more, uh, blurry, seeing a lot of sort of data and engineering teams work hand in hand to develop generative AI products.

I'm seeing that more than I've seen sort of data products in the past. Um, when it comes to observability, like you're, you're spot on. I think our thesis has been, um, you know, software engineers are expected to build reliable, secure, scalable products. And in the same way data teams, data and AI teams should be doing the same.

Um, by and large, we sort of see observability lagging infrastructure by a couple years. And so software engineers sort of took, you know, became, uh, you know, a practice methodology industry, whatnot, and then observability happened. And so, you know, companies like Datadog and New Relic and and whatnot, um, sort of fast follow that.

We saw the same happen with data. And so Monte Carlo was sort of really created, um, around the time when, um, cloud and data infrastructure solutions started, started to take, um, pace, uh, started to take, um, um, more and more momentum. Obviously companies like Snowflake and Databricks, um, and dd, DBT and Atlan.

Um, and with the, you know, more and more adoption of that also became the need to deliver reliable data, uh, reliable data products. And I think we're seeing the same thing with generative ai. The last couple years has been strong emphasis on the foundational models, the infrastructure, what does it even mean to build generative ai?

And I think folks are starting to realize that, you know, once you have a generative AI product, um, what does it mean to build a reliable product? What does it mean a reliable application? Um, so that's sort of, you know, from a trends perspective that we're seeing. Um, I would say, you know, another sort of, uh, maybe core fundamental difference, um, that I see between sort of data engineering and, and you know, what sort of the benefits of bringing some of the software engineering concepts to data are in how you think about building reliable systems.

So Tristan, you mentioned folks are not necessarily excited to think about data products or systems, but, but maybe they are. Uh, and, and you know, the long-term view there is a benefit to that. Um, one of the things that we spend time on as part of thinking about observability and reliability is. Thinking about your data state and all the reasons that something can go wrong.

And I think historically, if you thought about that five, 10 years ago, you really focused on the data itself alone. Um, you know, maybe the data was null or inaccurate or whatnot, um, that isn't sufficient anymore. And so data products, AI products can break for a number of reasons. It could be due to a problem in the data problem in a code.

The code could be code that's transforming the data, or it could be code that, you know, is used to build an agent or a prompt like, you know, changing that kinda logic can easily, um, you know, cause downstream impact. Um, it could be a, you know, a problem in the system, maybe, um, you know, a DBT job, uh, that failed or something else, um, some other transformation that failed.

Um, or it could be a problem in the model output. Like you can have the perfect prompt and the perfect context and the model output will still be inaccurate, uh, right. Hallucinations happen all the time. Um, and so, you know, I think when you start thinking about sort of introducing DevOps and software engineering concepts, it allows you to think about your data and AI state more holistically and consider all the different reasons for why and how to build reliable data and ai, um, products.

Awesome. A awesome answer. You know, what, what you just said reminds me of a company I I had talked to a few months ago, and they, like a lot of businesses, decided to spend some amount of time, uh, focusing only on, on gen ai, right? Basically took, you know, two quarters worth of engineering resources and said, just go to town and figure out what you can build with all of this.

And I, I think a lot of good things came out of it from sort of a product or, or AI product point of view. But when they got to the end, they realized this problem was uhoh. When a hallucination happens, did it happen because the data was correct and the model acted in some unusual anomalous way, or was the model doing the right thing and, and the data was, was wrong?

And that's a very hard question to answer unless you have the level of insight into both the data and the code that you're talking about. So that's, that makes total sense. Uh, PRPA for a question for you. So you, the, the, the cataloging industry has been around for a long time. It's been around, you know, pre cloud.

There's a lot of what I would call legacy data catalogs that have existed for a while. And I feel that what you're doing is you started to shift the conversation closer to the data itself. Um, sort of same question that I asked to Bar and Tristan, how, how has that been going and do you feel that there's still a further shift left, left to go and, and what, what does that shift look like?

Yeah. Um, yeah. So at Aline, I think we have this kind, like unique founding story where we were a data team ourselves. This was around the time when Snowflake and Databricks had maybe 12 customers each, um, so very already of days in the morning in data stack. And the way we were founded, we teams on these like large scale pro projects for data science, for social goods.

So I remember we, I was working on this project, which was India's national data platform, largest public sector data lake of its kind. Um, and we have this situation where we're implementing the entire data stack for customers. Um, and, uh, we would hit these problems that every data team faces every day, like, number on this dashboard is broken.

We don't know why. Um, analysts left on the team. I have once tried for two year, two hours because an analyst quit. And like, I was like, I don't know how I'm gonna tell you about this project to my customer because they just left with all the context, uh, about this project and this data. And so we tried to buy a solution.

At the time, nothing seemed to be right because, and I think we stumbled upon the reason that today that, you know, if you think about these words, data catalog, data governance, they have existed since the 1990s. And, uh, I have still not, uh, ever met a customer that has said, uh, data catalog and data governance works like, you know, I, it works inside my company.

Um, and our realization in that process was how disconnected this catalog governance was to the way data people actually work. Uh, and so all of this was a afterthought, like, let's document this data as an afterthought to data working. Let's put in a compliance check as an afterthought. Um, and it just never got integrated into the way data teams work.

And the reality is, data teams are dealing with breaking fixes every single day. It's a really hard job. So, you know, if you tell, like, the reality of being able to do an afterthought process just never happened. And so everything was outta sync. And then eventually your business users don't actually trust your, your data.

You know, they don't actually, you know, find like you, like, the biggest problem the data teams talk about is it's very hard for me to get business buy. It's very hard for me to get trust, like all of these, these problems. Uh, and it stems from this whole rule. And so for us, we sort of said, okay, then what does this look like if this was fundamentally built into the way data teams work?

Um, and we started this conversation of what, what we, we call embedded collaboration. So how do you, a part of that was how do you ship left? So how do you take these conversations into the way data people work? That's where, you know, data contracts, we integrate deeply into the way, you know, DBT and the way like people are shipping.

So that's one part of the, of the puzzle. The other part of the puzzle for us is actually also what we call shift, right? Uh, and shift right into the business workflow. Uh, and you know, if you think about this question of trust and quality, uh, trust data quality and trust is about data being fit for purpose, and purpose is defined by the business.

Uh, and so how do you build, build collaboration in both of those angles, uh, has been our biggest differentiating point today, um, as we see the market evolve and just say, okay, how do we build this new form of governance and cataloging in the ecosystem? Great answer. That's awesome. Um, okay, so I, I wanna this, I've got a bunch of questions for you guys.

Um, some of them came from a, a panel we had a little bit earlier in the day, um, with, with some investors. And, um, Chris, Chris Riccomini, who's a, a friend of mine, and he is also a, an author, um, I think made a pretty lucid point, which was that right now the state of, uh, data software development is, is reasonably fractured, right?

You, you have the analytics engineer, which of course, sort of DBT kind of pioneered. You've got the data engineer, you've got the BI engineer, which I'm, I sometimes am not exactly sure what that role is. Then you've got sort of software engineers who are doing work with data, and his view was that at some point these roles need to come together to form a bit of a full stack data engineer, if you will, in the same way that, you know, you've got front end and, and backend engineers that can kind of double up and, and, um, do different tasks.

Uh, I, I would love to know what your opinion on this is. So like, what is the future of doing data development and what are the roles that these different sort of cuts of engineering end up playing in relation to, you know, who's writing the transformations, who's managing the data quality, who's setting up the catalog, who's understanding the lineage?

Like where does that responsibility end up falling?

Who do you wanna start? Anyone you can start since you spoke. I, I have, uh, I have a lot of thoughts on this. Um, I think the big, the big part of this question that you didn't, uh, mention is ai. Um, I think that, I wrote about this recently. I, I think that AI is, uh, coming for what we currently think of as the data engineer.

And that doesn't mean that there's no work to do in data engineering. It means that the package of tasks that we currently think of as the responsibility of the data engineer is going to be shifted around. Um, I think that if, if you look at, uh, what data engineers spend their time on, uh, that. A large majority of the, the time, the task time of a data engineer today in a healthy data infrastructure, uh, can, can be automated with current generation, large language models, as long as you've fed them the correct context.

Um, now there's, there's productized experiences that need to happen in order to like, make this work better and kind of make a workflow around it. But fundamentally, like we, we don't need any like, massive improvements to, uh, the state of the art from a model technology perspective to make that happen. So you like have this explosion, like, okay, take this set of tasks and explode them, and who, where, where does it all go to?

I, I think that, you know, we're gonna continue to call people who build applications, software engineers. We should continue to do that. Great. Uh, we, uh, are going to have more folks than ever I think move to the data platform level, uh, because we need great data platforms for all this to work on top of. Um, and then there's gonna be more humans, uh, than than ever that do two different things.

One is like, are close to the business logic and can use AI and high leverage tools to actually build the, the pipelines, uh, in, in this kind of data platform. But then also, uh, kind of business application developers that, that like work on automations because automation, there's like more scope than ever for, for automation.

Um, but, but I think that the differentiation between like data engineer and analytics engineer and data analyst and bi, like all of that stuff is kind of going away because they previously there, there was like some mapping between those roles and the tools that they used. And increasingly all the tools are accessible to all the different personas.

And I could say a lot more about that, but like, I, I don't think those, they all provide kind of the same value to the business and they're all being acted upon in kind of the same way. Sorry, it was a lot. Uh, bar prpa. Any additional thoughts on that? Yeah, I would second what Tristan said. I think anyone who is doing their current job and data, who thinks that they're going to be doing the same job in the next couple of years is, is already behind.

I think, uh, the, the question I think all humans need to answer is, what is the job that humans will continue to do that AI will not take away? And the answer to that, at least in my thought process today, is it's going to be closer to determining purpose and business and how business wants to operate.

Like getting much closer to that. Much away from the actual doing the thing, the actual writing the code. So yes, I agree. Platform will continue to be super important. 'cause that's going to enable everything. Uh, the role of the analyst I think is honestly a question I, I already know. I was talking a CEO last week and they've cut 25% of their analyst team and they're refocusing like that budget on like, so it's, I think that the, that role like translating what the business needs to someone else to write code, like that job I think just like goes away.

'cause language changes it. Like that's the most beautiful thing that LNS have done. It's language, it creates, it makes it easier for that conversation to happen. So in my head, it feels like it's, i I think for the next five years it, it's likely to be some combination of a product manager type role and a engineering type role.

And it's probably going to be full stack. So the product manager type role, data, product manager, whatever you decide to call it, closer to business, closer to understanding what needs to happen. A data engineer ish type role that is, that is the, I still think is going to be more about the, or closer to technical.

Um, and there's probably a future in the next five years where both of those roles also emerge into just one role. Um, that's my general sense.

Maybe just for fun. Oh, go for it. No, no, no. Please. No, I was gonna say maybe just, I wanna hear what's gonna be fun. Just for fun. I'll, I'll take the contrarian argument. Um, you know, I, I'll, I'll maybe I'll make the case. You know, just we talked about comparisons to software engineering. I thought, I think software eng, how we write code and what we do, how we build software applications is, has changed tremendously.

And, and yet there's, there's a lot more engineers, not less, you know, it's perhaps the, the, the rise of, um, the rise of the engineering role in, in, you know, the last couple of decades, probably maybe the fastest growing role other than data scientist, I think. And so I think all these advents in AI in particular will actually increase the need.

Perhaps the jobs that we will be doing are different and I think AI can make teams a lot more productive. I, I certainly think, you know, great teams see like 10, 20, 40% improvement in productivity, um, very, very quickly, especially for, um, for less experienced teams in particular. Like, it helps kind of get up to speed.

Um, I think for, uh, but I, but I don't think the, the jobs are going away. I think the, actually the importance of all these people, um, and all these roles will, will be bigger as a result is more important. Um, I also think. Roles that are closer to the business will continue to be more important because the ability to understand data and understand the meaning of it.

It's interesting, you know, we still see, um, you know, I was, I was talking to this, uh, head of a ai, like a large Fortune 500 company a couple weeks ago, and I was asking him like, how do you guys like, make sure that, you know, the agents that you're building are actually like, you know, doing something that's not crazy or, you know, um, and he was like, oh, we have, you have users just sifting through hundreds of responses every single day to make sure that it's not crazy.

And so that reminds me of like five, 10 years ago when we were like sift through data and reports to make sure they're accurate. Um, you know, the, the role of human in the loop is, is still very critical, I think, especially for critical decisions and, um, you know, um, uh, performance of, of specific functions.

Um, so I'll, I'll take the, I'll take the devil's advocate here. Uh, and, and claim for, for the contrary. Yeah. Okay. Uh, I, I love that. Great. If, if there's multiple people and, uh, we all completely, uh, agree about what's gonna happen in the future, there's, uh, a hundred percent chance that all of us are gonna be wrong.

If some people disagree, then at least there's some percent chance that a few of us are gonna be right. So, um, well, I also think, I think there's like, just to, I think there's a difference between the job is gonna change drastically and the job is going to go away. So I think what Bar is saying is the job, like data is gonna become more important, is a fairly truthful statement, which means there will be more humans that will probably enable us to finally hit the AI native company data.

Like, you know, there's, there's this, so I think it just a job is gonna look drastically different than it is gonna look today. So I actually don't know if we're debating that. I Yeah, yeah. I I don't see it as much of a disagreement. I see it as like, yeah. Uh, I, my statement was not that there's gonna be fewer humans.

It is that there's gonna be a massive shuffling of jobs to be done and job titles Fair enough. And I think that's almost certainly going to be the case. Um, okay. So not, not to change gears, but maybe to sort of, uh, continue to elaborate on, on the same set of points. Um, when I, I think the sort of modern data stack really rose to prominence and Snowflake really rose to prominence.

The, the primary use case for data at that point in time was still analytics. And over the last decade, I think there's started to become this, this shift, uh, where you, you started moving more into, uh, machine learning and then you had sort of these, uh, you know, operational use cases for data. And I know that Snowflake is now pushing data apps, so can you like build products on top of Snowflake?

They're doing all types of interesting production things. And now of course you have have ai. Uh, maybe as sort of the last, the last question of the day, um, how are you all, number one, do you agree with me? Do you think that that sort of transition is happening and is it having an effect on the way that teams are doing their work and the level of importance that executives are starting to place on the, the infrastructure to, to manage data?

Yeah, I, I can start with this one. Um, so if you go back to the premise of data, like why this, like, you know, data democratization, self servers, all of these words that have been buzzwords. I think for a long time the premise started at we need to convince, we need, convince is the wrong word, but we need to get humans to use data better, to make better decisions because our business is going to get better.

And ultimately the premise was humans need to drive action based on this data. Uh, and I think the shift that has happened over the last five years is a question on what drives action? Who drives action and what is the role of the human in driving action? So the first shift, I think, and this started happening I think four or five years ago, where operational systems started becoming a lot more important, right?

And that started becoming things like, if you think about running a automated marketing or a marketing campaign, like before it used to be a human picking up the list, making a decision. Can you start making that like a real time system? And that was the first operational shift that happened to say, okay, what types of positions can become actually completely autonomous or, you know, in, in the way that they're running.

I think that shift is moving even more. Like the question is like, you know, uh, I had a customer the other day and I said, you know, we, I had this thing where I said, trust in data, trust in the model and trust in the ai. And he pushed back on me and he said, uh, actually the last step alpa, that you miss this trust, trust in the action, not just in the ai because we are going to have our, like actions in our business are going to be done autonomously.

Uh, and I think that shift has, has made the role of infrastructure, data, model trust, all of these other things behind the scenes a lot more important. Uh, because the way we, like, if you think about how do we run a great business, you can go on to say, I think you can go on to say that only AI native companies will continue to be success, like exist in the next 10 to 15 years.

And infrastructure is going to play the role to make that action possible. And so the way the same thing needs to be accomplished five years ago to today has changed drastically.

Great answer. Borrow Tristan?

Uh, sorry, just I want, can you, can you clarify like what is the shift that you're talking about? You mentioned Snowflake is pushing this, but I'm not sure I understand what they were pushing. Yeah, so, uh, snowflake for the last few years has this concept, I'm sure you've heard of it, of like a, a data application.

See, and the idea is, oh, see, right? Like, you know, before you, you had sort of your, sort of your, your prototypical software apps, they would generate some data, it lands in Snowflake, and then boom, you build a report on top of it, and that's cool. But what happens when all this useful data that's sitting in Snowflake, you want to repurpose it into some other operational thing that a customer is interacting with, which I guess could be an embedded dashboard, but it could be some other workflow automation based thing.

And like, that's, that's like different, like it's, it's gone from, um, hey, my, the numbers in my report look off to, hey, like there's something that's failing and it might be costing me real money. Got it. Yeah. I mean, look, my perspective is that players like Snowflake, Databricks, Amazon, Azure, et cetera, like these platform players, they will always wanna own more and more of it.

Um, and, and that's okay. That's good, right? That, and so I think you're seeing them sort of try to do more, um, horizontally and vertically, and so they're trying to own the whole thing, the whole pipeline, right? I think that's like one trend that you're seeing. On the other hand, if you look at a common enterprise, they're very likely to have a, a very non homogenous estate.

So they very often have like, you know, multiple data, data, lake houses, multiple data warehouses. Um, oftentimes many of them are actually still on-prem and, and their big push is to move to the cloud. You know, I think we like to talk a big game of generative ai, but you know, a lot of folks still need to move to the cloud first.

Like doing gen AI on-prem is, is, is harder. Um, and so like, I think the reality is for enterprises, there's a long way to go. Um, and it's also very likely that they will make multiple bets, um, with different infrastructure players. I also see that in foundational models. So every single day there's a new model and we're seeing this like, you know, battle of the giants between open AI and thero and, and whatnot and open source and closed source.

But very not uncommon for an enterprise actually, you know, not common for enterprises to switch models every week, in fact. Um, and so I think, you know, if, if I am a, you know, chief data officer or CTO, like how do you, how do you work in that, in that, in this ever-changing very fast world? I think there's a couple of things.

One is I think it's okay to, to hedge your. You're hedge your bets and, you know, make a couple of, um, you know, just a couple of players to, to align with. Um, I think that'll continue to be the case. And no matter what the, you know, no matter what bets you make, thinking about, you know, obviously the, the trust and reliability and the whole estate and the output is critical.

Um, and I think that doesn't matter, you know, which infrastructure player you choose or who you go with or what their strategy is. I do think it's a tricky time, you know, sort of figuring out like who to go with as I, I have empathy for, uh, you know, for data teams trying to, to place their bets and trying to stay with the, the latest and greatest.

Um, I will say I do think, you know, access to the latest and greatest model is something that we all have, you know, uh, within a couple of seconds or a couple of minutes, I can, uh, very quickly get, uh, you know, an API key and, and stand that up. And so, um, that actually is not the advantage that many enterprises have.

In fact, it's the reliability of their overall systems, the reliability of their overall data, the foundation that they're building on. Um, in my, in, in my view, that is really sort of the, uh, you know, the, the moat because degenerative AI application is only as good as the foundation and the data, um, that is powering it.

Uh, those are my thoughts. That's awesome. Tristan. Uh, last word. Oh, I had a really smart thing to say, but I see dark timekeeper hopped back on video and so I know, I know my time is, IM the music at the Oscars, you see me? Yeah. My, my short answer is like the lake, the, the, the, the data lake is, is everything.

And this is a story that's been going on for a lot longer than, uh, you know, the idea that like the data infrastructure would be powering applications is, is not new and uh, the lake is at the center of all of it. Awesome. Well everybody thank you so much. Um, this was a great panel. Uh, we really appreciate you having here and uh, we will, we will see you Elise, you'll see me again very soon.

Can't get rid of you. Keep coming back. Well, that was awesome. We had some semis spicy takes and I think the chat was loving it. I also feel like we can create this in a debate format next time so we can have different sides debating out their, uh, what they think the future holds. However, as I mentioned before, we gotta keep this train moving."