In this #MOEexperts conversation from theCUBE + NYSE Wired at the NYSE CUBE Studios, theCUBE’s John Furrier sits down with Jake Graham, co-founder and CEO of Bobsled, to explore how data infrastructure is becoming the backbone of agentic AI and enterprise value creation. Graham explains how Bobsled – described as the most adopted data sharing platform – helps the world’s largest data producers (including major stock exchanges and leading data providers) build, distribute and govern interoperable data products across wherever consumers work. The discussion connects Wall Street and Silicon Valley priorities: making high-value data understandable, accessible and governable across Snowflake, Databricks, S3/Iceberg and more, while enabling emerging access patterns like MCP integrations into Claude. Graham unpacks why text-to-SQL has proven harder than expected, arguing that “context engineering” and an interoperable context layer – rooted in knowledge graph principles – are crucial for agents to route and interpret enterprise data correctly.
The interview also highlights notable business and market signals: Bobsled is a three-year-old, consumption-based business that went fully live ~18 months ago, raised its Series A from Madrona, 406 Ventures and Greycroft, grew revenue 5x last year and has more than doubled this year – the team is ~30 people after 10x headcount growth in the past 18 months. Graham details how Bobsled governs data footprints across Snowflake, Databricks, BigQuery, Redshift, Microsoft Fabric and open table formats and why pushing metadata “down” (e.g., S3 Tables) matters for flexibility and latency. He shares how Bobsled underpins data sharing in top cloud marketplaces and why openness (vs. walled gardens) will let platforms like Salesforce, SAP, Workday and ServiceNow blend their data with customers’ estates to power agentic workflows. Key topics to listen for include: real-world use cases for data productization, evolving partner ecosystems, the rise of “SRE for data,” marketplace interoperability, MCP and A2A dynamics and forward-looking takes on where the “intelligence layer” should live as finance and tech converge to shape enterprise growth strategies.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Mixture of Experts Series
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Mixture of Experts Series.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Mixture of Experts Series
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Mixture of Experts Series. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Jake Graham, Bobsled
In this #MOEexperts conversation from theCUBE + NYSE Wired at the NYSE CUBE Studios, theCUBE’s John Furrier sits down with Jake Graham, co-founder and CEO of Bobsled, to explore how data infrastructure is becoming the backbone of agentic AI and enterprise value creation. Graham explains how Bobsled – described as the most adopted data sharing platform – helps the world’s largest data producers (including major stock exchanges and leading data providers) build, distribute and govern interoperable data products across wherever consumers work. The discussion connects Wall Street and Silicon Valley priorities: making high-value data understandable, accessible and governable across Snowflake, Databricks, S3/Iceberg and more, while enabling emerging access patterns like MCP integrations into Claude. Graham unpacks why text-to-SQL has proven harder than expected, arguing that “context engineering” and an interoperable context layer – rooted in knowledge graph principles – are crucial for agents to route and interpret enterprise data correctly.
The interview also highlights notable business and market signals: Bobsled is a three-year-old, consumption-based business that went fully live ~18 months ago, raised its Series A from Madrona, 406 Ventures and Greycroft, grew revenue 5x last year and has more than doubled this year – the team is ~30 people after 10x headcount growth in the past 18 months. Graham details how Bobsled governs data footprints across Snowflake, Databricks, BigQuery, Redshift, Microsoft Fabric and open table formats and why pushing metadata “down” (e.g., S3 Tables) matters for flexibility and latency. He shares how Bobsled underpins data sharing in top cloud marketplaces and why openness (vs. walled gardens) will let platforms like Salesforce, SAP, Workday and ServiceNow blend their data with customers’ estates to power agentic workflows. Key topics to listen for include: real-world use cases for data productization, evolving partner ecosystems, the rise of “SRE for data,” marketplace interoperability, MCP and A2A dynamics and forward-looking takes on where the “intelligence layer” should live as finance and tech converge to shape enterprise growth strategies.
In this #MOEexperts conversation from theCUBE + NYSE Wired at the NYSE CUBE Studios, theCUBE’s John Furrier sits down with Jake Graham, co-founder and CEO of Bobsled, to explore how data infrastructure is becoming the backbone of agentic AI and enterprise value creation. Graham explains how Bobsled – described as the most adopted data sharing platform – helps the world’s largest data producers (including major stock exchanges and leading data providers) build, distribute and govern interoperable data products across wherever consumers work. The discussion con...Read more
exploreKeep Exploring
What is the role of Jake Graham and what does Bobsled do in the context of data sharing and products?add
What is the most adopted data sharing platform on the market today and who are its primary customers?add
What are the key requirements for making data available to enterprises globally?add
What are the key challenges in creating and managing data products at scale?add
What aspects of data engineering and governance are being discussed in relation to the infrastructure and tools being set up?add
What are some current developments in AWS related to data and metadata management, and how do they compare to other platforms?add
What are the key components that contribute to the effectiveness of Bobsled's services in managing infrastructure and data across multiple platforms?add
>> Hello, I'm John Furrier with theCUBE. We are here at our New York Stock Exchange CUBE Studios. Of course, we've got our Palo Alto, so connecting Wall Street and Silicon Valley, tech and money. Of course, this is our mixture of experts series that we'd like to bring all the experts in who are making things happen. Data is the key to AI and that's one of the big focus areas we're seeing in the market domain expertise and having data available for what's coming to be the most hyped up area will soon be the agent and infrastructure. Jake Graham is the Co-Founder and CEO of Bobsled. They're doing innovative work around making data work and getting value out of the data, which is the number one priority. Jake, thanks for coming on theCUBE. Appreciate it.
Jake Graham
>> My pleasure. It's great to be here.>> So you guys are working in an area that we'd love. We cover it pretty deeply. The Databricks, the Snowflake. The data cloud was the big wave first, basically data warehouse in the cloud. Okay, get that. Now, the focus is I got data, I got to make it work for my business. So explain what you guys do and then we can get into some of the things you guys are working on in terms of the market and the products.
Jake Graham
>> Yeah, that'd be great. So Bobsled is the most adopted data sharing platform on the market today. Our customers use us to build, distribute, and govern data products across any platform in which the consumers of those data products might work. So we primarily work with the largest data producers in the world. So whether that's three of the five largest stock exchanges, whether that's companies like Dun & Bradstreet, or ZoomInfo, or Cotality, or many others, companies whose core offering really is data have an existential requirement to make sure that data is understandable, accessible, and governable really across every platform that sets in existence today, which really creates a really complex infrastructure problem, which we solve for them.>> So you guys target people that have a lot of data. They're data full as we say. Enterprise is becoming more and more data full as they realize they've got to store more. With AI, they're storing everything. What specific market are you guys targeting? Explain the approach, because there's the analytics market, you've got the GenAI market, those are the two hot markets. You've got Databricks, you've got Snowflake, you have a variety of other products. What are you guys targeting? What's the market?
Jake Graham
>> So we're used today ... I think there's many ways to think about the data market. So we're used by, again, in financial services and healthcare and technology, but our primary customers today really are data producers. It's core to what they do. So they aggregate structure and distribute data to their consumers. So whether that's companies like the New York Stock Exchange or we work with Deutsche Börse or major stock exchanges that need to distribute all of the data that they generate to every financial institution in the world, or whether that's a company like ZoomInfo or Dun & Bradstreet who are structuring data around just about every company that exists. Our primary customers really produce and store the most unique and valuable data sets in the world.>> So do you cover the enterprise data, or is it more the data businesses?
Jake Graham
>> So we're across both. I mean, I think our largest users are creating and structuring enterprise data, and then we also work with via these data sharing networks, the largest institutions and enterprises in the world. So whether that's just about every bulge bracket bank, vast majority of healthcare institutions, major technology organizations, just about everyone touches Bobsled's data sharing network in some way, but our core users are those data producers.>> So what's the core problem that you're solving? Because obviously, data is core to those businesses and obviously the large scale enterprises too. What's the problem that you guys solve?
Jake Graham
>> So in order to make data available to enterprises across the world, there's really three things that you need to do. You need to structure your data in a way that's understandable. And this concept of a data product is one that Bobsled has been leading on, but is really starting to be understood across the market, which the way I think of it is you need to be able to have the core data itself and the business context that describes it. So the first problem we solve is how do we create these concepts and manage at scale the idea of data products. The second thing that needs to happen is that those data products need to be accessible and consumable wherever your customers work. So one of the things that I really believe is that the concept of data gravity, of everyone bringing data to them and storing it in one place, it's really broken. There's not one platform, there's not one owner, there's not one silo of data. Data needs to be able to flow into the different platforms in which it's leveraged. So a major reason our customers use this is their consumers are in Snowflake and in Deutsche Börse, and in S3, and they're in Iceberg. And now increasingly, they're leveraging MCP integrations into Claude. So we are an interoperability layer for those data products.>> So your main purpose, if I get this right, is to get value out of the data.
Jake Graham
>> It's to make data accessible to generate value. Exactly.>> Got it. Okay. So how about the business? How old is the company? Why were you formed? What's the status?
Jake Graham
>> Yeah.>> Take us through the business model. How does it work?
Jake Graham
>> Yeah, absolutely. So we've been around for just over three years. In terms of status, we raised our series A not too long ago from Madrona, 406 Ventures and Greycroft who have been great partners. We really fully went live and in production I'd say about 18 months ago. And in that time, we've scaled really quickly to a significant portion of the world's largest data companies. I mean from a revenue perspective, we grew 5x last year. We've more than doubled this year, and I think we're getting to a pretty significant percentage of data flows. Again, touching Bobsled in some way. In terms of our business model, we're all consumption based. It's based on the number of consumers and the number that's actually flowing through Bobsled. And I mean, it's been a lot of fun to be building this at a time that we're going through the largest platform shift at least since the internet. And I think we've got a really front row seat into how are the access patterns on some of the world's most valuable data sets changing, and how is the value from structured shifting to semi-structured and unstructured data as well? How are access patterns changing with capabilities like text to SQL, like model context protocol? I think it's been great to be building in a time in which the rate of change is so high that you really just have to get in and listen, but make bets on where the market is going. It's a lot of fun.>> You're an infrastructure data, infrastructure play.
Jake Graham
>> Exactly.>> I mean, Madrona is a great firm, great pedigree of investors. They invest in technology companies. We know them very well. Love Madrona. So congratulations on that.
Jake Graham
>> Thanks so much.>> But they invest in tech, so I can see that happening. And also, the demand side with GenAI and analytics is there. So analytics has been around for a while. We've seen that, but the GenAI is putting massive pressure on the infrastructure to be scalable and reliable. You mentioned MCP. One of the greatest things this year, I think organically it's grown is MCPs. It's still growing, security concerns and other things.
Jake Graham
>> Very early.>> Still early, but it's a good sign.
Jake Graham
>> Yeah.>> The agents are coming. I mean it is hyped up, no doubt, but you're doing the work to make the agents work. How would you feel? What's your take on what's the readiness factor? What are the requirements for agentic? Because agentic infrastructure is data infrastructure.
Jake Graham
>> Yeah, absolutely. I mean, I think it's been really interesting to see that one of the use cases that I think we all expected to be one of the first and easiest implementations of agentic AI in the enterprise, the idea of text to SQL, how can I chat with my database has actually proven to be one of the hardest ones. I think it's a really interesting reason why, and this is where I'm spending most of my time now. The ability to take language and turn it into a SQL query is actually relatively straightforward. What's proven to be really challenging is to get the model to actually understand the data that it's interacting with. Because while data, structured data is well-built into tables or in platforms that have got an enormous amount of metadata, what it lacks is the business context. So what we're seeing in practice today where agents are really hitting the wall is when I have a user who is asking for information around revenue, for example, there might be 75 different definitions of revenue that are not mutually exclusive, and it becomes incredibly challenging for an agent to be able to route that data correctly. And the answer that seems to be becoming clear, but that we really need to build the infrastructure around is this idea of the context layer. And what we really bring is interoperable context layers for the world's most valuable data products. One of the things that I talk about a lot is that I think over the last ... We saw from over the last 10 years, data science, but then increasingly data engineering became one of the most critical concepts within the technology landscape. As we move from as an application focused technology market to a data focused market where we're now moving to this model focused market, it's actually context engineering that I think is becoming the next wave and making agents work. So a lot of the infrastructure that we're building is how do I, using the frame of a data product, take all of this data that exists, but associate what is the business logic around this? How is this data used? Who uses it? And how do we help agents actually understand how to interact with structured data?>> It's very interesting. I like where you're going with that because a lot of the conversations we have, you have a couple of different perspectives, the semantic layer, the harmonization layer, and then you have that platform engineering world where it's like ex-SREs at Google, they think differently. They think like platform, not so much semantic layer, but there's a mix, because you want the context. You got to understand that piece of it as well as the platform engineering mindset. How do you look at that? Because there is a spectrum, but there's also a middle ground between both data engineering. I mean, even Amazon, SageMaker and Bedrock are two different animals.
Jake Graham
>> Very much so.>> One is plumbing driven, and one is model garden, whatever they call it, model selection for developers.
Jake Graham
>> Yep, and we use both. And we use actually all the platforms you're talking about.>> Explain this nuance, because data engineering is becoming a skill.
Jake Graham
>> Yeah. You'd ask what customers we focus on. One of the things that I'm most excited about in this market is all of the building blocks to solve these problems have existed for years, but what's really lacked is the incentive to invest in them and make them real. So the idea of a semantic model was just previously called a metrics layer. And my background actually is in knowledge graph, so 10 years ago, we were talking about ontologies and RDFs moving into property graphs. All of this infrastructure has existed. What you haven't had is the right people to be incentivized to actually implement them. So the reason we started with data producers, data companies is they really have product management principles around there are people who get promoted if a table is consumed more in the market. And that's really what you need, because a semantic model, it starts really simple. I need column level descriptions of data and then you realize, "Well, actually, I need a data dictionary as well." And every platform represents those in different ways. And then you realize, "Well, actually, I do at least need some form of ontology that maps industry language to how I represent this data." And then you realize, "Well, actually, I do also need to be able to give sample prompts and sample SQL queries and sample Python notebooks."
And then for our customers to be able to say, "Well, actually, we power six other platforms and we have support tickets." So we started with the companies that are most incentivized to really construct and bring these data products to life. What I'm most excited about for us as a business is that's quickly turning into every enterprise. I mean, as someone who's ... Product management has been my craft for so long, it's really exciting to see that start to come to data. And really, what Bobsled is focused on is creating this data product management layer that will work for all of the world's most valuable data assets to be consumed.>> And you're seeing data marketplaces, so productization of data is a real deal.
Jake Graham
>> So we actually are, in some cases, the largest but the top five largest manager of data sharing in every data marketplace. So whether that's Snowflake, or Google, or AWS, or Databricks, we're closely partnered with all of them. And if you look at a lot of the data partners of the year of these companies, many of them leverage Bobsled to actually interact with these.>> Yeah, you guys are under the covers. Hiding in plain sight.
Jake Graham
>> Yeah. That's the idea. That's the idea. It's excited to come to the show, up to the world here.>> Hey, theCUBE is doing our part to unveil all the action. You mentioned knowledge graphs. I had Sudhir on, who's now the President and CPO of Neo4j where you have history and we were talking about the web 1.0, Google Search, contextual behavioral data, the two real key flywheel elements of it. And then I said, obviously, agents for data. He actually interjected and said, "No, no, Google then went to knowledge graphs." So it was a nice call-up, but my point is that knowledge graphs have been around at scale for Googles of the world, but not everyone could be Google, because they had a unique situation with their data centers and what they had. But now, everyone is looking a lot like Google and all the top use cases and data right now are RAG and search-like paradigms.
Jake Graham
>> Absolutely.>> And generative AI is essentially a prompt. It's a search.
Jake Graham
>> That is absolutely, totally.>> So we are getting into an era of real-time runtime assembly of data. I mean, lack of a better description, I don't know what to call it, but generative.
Jake Graham
>> Yes.>> That's what's happening, but not everyone is ready. How are you guys helping companies get ready? Some people are like just scratching their heads, "Okay, I have a lot of data, it has value. How do I figure it out?" I can see that clearly. What other use cases you guys working on that you could share that would illustrate some of the things you're working on for your customers?
Jake Graham
>> I really appreciate that question and you're really just speaking to my background. I was at a knowledge graph company that Intel acquired called Saffron, and then I actually left to join Neo4j and managed all of their knowledge graph and graph data science applications before eventually founding Bobsled. I'll call out Emil as an investor in Bobsled.>> It was great.
Jake Graham
>> So I'm bring that up. So I think there are two major problems that we are really making inroads to solve. So repeating, again, the first one is finding the people who are incentivized to bring the context, which at the end of the day is a knowledge graph to data. We're giving this context engineering interface to be able to say, "This is not just a table or a set of files, it has business context around that." And so that's kind of the manual integration process for how do we give the interface a human to input all of the context around the data product, but that only really becomes valuable if that knowledge graph, that knowledge representation of that product can evolve with usage. So the next thing we're doing is making sure as you're creating agentic and prompt interfaces in addition to the existing SQL and Python and other interfaces, how do we make sure that you have an automated feedback loop that allows for an agent to understand when someone is asking this question? This is what they mean and this is the data to point at and this is how to interpret it. Because I think the way the world is going, a huge differentiating aspect for every enterprise for the last 10 years has been around the ability to generate insight from data, which has been a human bottleneck. Do you have the right data scientists and data professionals that can get insight from data? And now what we're getting is that ability to generate insight is being commoditized, right? The ability to take a question and leverage data and that interpretability is becoming really quickly commoditized. What's really interesting from that is that it makes access to unique and valuable data the real differentiating factor, but the gap we have to cross is that you have to be able to associate that business context with that data and you have to do it in a way that learns. So really so much of what we're focused on is working with these companies that have these not just valuable data products, but the incentive and the people and the history.>> Connect the data, incentivize them, monetize.
Jake Graham
>> Exactly.>> So we're seeing the data plumbing, and for lack of a better description, come out. I said on theCUBE I think years ago that the Kubernetes cloud native developer world, which is DevSecOps, we all know, was going to move to data.
Jake Graham
>> Absolutely.>> And I said, and I'd love to get your reaction to this, "There'll be an SRE for data." SRE was Google's term for site reliability engineer that managed a fleet of servers. And Dave Vellante and I were debating on theCUBE Pod and the rationale was data will have scale like servers and you get consolidate it and have one person manage a fleet of data sets.
Jake Graham
>> Absolutely.>> You agree?
Jake Graham
>> Very much so.>> And that's kind of what you guys are setting up?
Jake Graham
>> Very much so.>> The tooling around it, the infrastructure?
Jake Graham
>> Yeah, I think it's the ability ... I think there's two layers that are happening here. One, so the concept of SRE is, and I very much agree with it, is moving into data. And I think data engineering is a very broad term, but so much of the role of a data engineer is to actually understand the network of footprint of where our sources, how is data transformed, where is it consumed and what we consider to be data governance. So that's a major part of what we do. What we are the best platform in the world for bar none is the ability to do that interoperably across every data platform. So if you need to manage data across Snowflake, Databricks, BigQuery, Redshift, Microsoft Fabric, open table formats across every cloud, we give you the ability to govern that footprint. Where I think it's going now is, again, it's the ability to also associate and understand the context around that. Because today, there's really a bifurcation of you either are working on the pipeline, you're a data engineer, or you're working on the insight, you're a data scientist or you're a data analyst. And there's no one that really sits in between to be able to help bring those things together at scale. That's really what Bobsled is doing with this data product frame.>> I think that validation has been, and I brought this up earlier and I think it's important to illustrate it, Amazon thought SageMaker would be the end-all be-all for this layer. And then when OpenAI came out, Bedrock became more of the developer interface, because they just had needs, and you mentioned data science, so certain personas want to access certain things. And then the nerds who were plumbing and tweaking the instances and all the hardware were optimizing using SageMaker to handle kind of that platform piece. Now, it's basically the same thing, but one is tailored towards more knobs and buttons to turn for configuration, for performance, management, a little bit different skill sets. So again, the skills between those two roles change. What's your reaction to that piece? Do you agree? And how does that translate into your customer base? Because you mentioned a data scientist, they care about what they work on. They may not care about is this optimized for routing context for an LLM router through whatever or managing graph database. I mean there's different worlds.
Jake Graham
>> So two thoughts. I think the role of a data scientist is going to shift enormously over the next few years and I think it's going to move ... You'll still have cases in which many data scientists are working on highly complex, I think more mathematically advanced problems. So much of the work of data scientists today I consider to be advanced analytics. How do we help business users ask questions and make predictions against data? I think data scientists are going to be asking less questions and providing more context for people to ask questions themselves. When we get into AWS mean, this is just such a fun space to be in right now because we're all trying, we're all grokking and wrapping our minds around the same problems and there's many different ways that they're being solved. I think one of the most interesting things that AWS is doing right now around data is actually pushing metadata down into ... So, for example, S3 Tables.>> Oh, S3 Tables, phenomenal. They got vector database built into S3.
Jake Graham
>> And then you've also just have this concept of the data catalog has never been as exciting as a topic as it is right now. But what is a data catalog? To your point, who is the user? Is it a data scientist? Is it a data engineer? Is it an analyst? And does that metadata live like AWS believes actually at the object level? Is it something that I think more towards Microsoft's belief that sits above that in OneLake? Is it something that should be existing apart in Alation or Collibra? Bobsled has had to build a very particular type of data product catalog to work across interoperability. It's a space that's growing really, really quickly.>> I mean you mentioned RDF and the other things you worked on. I would say that from what I hear and see is that pushing metadata down lower is better because your faster latency, better latency on finding something. I mean, a lot of this indexing technology has shifted. So that's a whole nother-
Jake Graham
>> Yeah. I mean, so much of these shifts as well come with kind of unbundling of platforms as well. And I do think there's a tension there, because the more that you rely on one platform for the more ease of use you get generally. And it's much easier to say, "I'm going to do everything in one particular platform and kind of abstract away all of this metadata management." But I think what's happening is this data is scaling significantly and is moving from being data was exhaust and then data was oil, and I'd say data now is gold. As we're moving to this really data-focused technology landscape, interoperability is becoming much more important. And when you do that, the more you can push down the metadata, the more effectively you're going to be able to swap out engines, you're going to be able to swap out governance, you're going to be able to swap out interaction. And I think that tension, it moves back and forth of, are you in a Cambrian explosion or a consolidation moment? I think what's interesting is we had that Cambrian explosion with the modern data stack. I think we saw quickly consolidation, which I think will still happen to a degree, but I think what we're doing is we're actually moving back to another explosion because of this platform shift. And that means we need standards, we need to push things down.>> And these guys are really getting into the operational side of things.
Jake Graham
>> Exactly.>> Because the operational analytical database has been always been the great divide, right? The two worlds-
Jake Graham
>> Absolutely.... >> are two now coming together.
Jake Graham
>> The HTAP database debate. One day, there will be a successful HTAP database.>> Talk about the secret sauce that you guys have. What is the secret sauce? Obviously, Madrona invests in IP. What are some of the things you guys have? What's the calling card for Bobsled? What's the secret sauce?
Jake Graham
>> I appreciate that question. I mean, at a high level, Bobsled exists to provide interoperability across data structures. So I mean, that has evolved into a few major pieces of secret sauce, I'd say. The one is infrastructure management itself. I think the core reason the cloud exists is the part of the core reason we exist. Managing infrastructure is burdensome. It looks the same across just about every company.>> Grinding things out, I mean it's like-
Jake Graham
>> Having an abstraction layer is fantastic. There is no platform better for Bobsled, period, today. If you need to run data platforms across every cloud in every region, we do that incredibly well. I think the second layer is around data structure and context interoperability. So as we're talking about, it's great to have standards and push things down, but those don't exist today. So our customers, as a rule, are not in one platform, they're not in one cloud, they're not even entirely in the cloud. They're not one data platform. So giving an abstraction layer to say, "This is a schema, this is how data is optimized, this is how data updates." But I think more importantly, here is the business logic and context and usage around that, I think we've done a great job at. And then in addition, the access layer. So one of the reasons I actually founded Bobsled was I was at Microsoft and I was looking at a lot of the innovations not only we were working on, but I think Snowflake sharing really was one of the originators around and we've seen so much evolution there. But those cloud-native access mechanisms of Snowflake sharing and Databricks sharing and BigQuery sharing, none of them speak to each other. So if you want to be able to leverage interoperable access mechanisms, Bobsled is again the best place to do that.>> I mean, I love the abstraction layer and I love the interaction access layer, because it brings up the idea of what agents will be doing, right? Because when you have abstraction, you can abstract away complexity.
Jake Graham
>> Exactly.>> And make stuff work under the covers as they say. So I have to ask you companies like Salesforce, one company I've been very critical of over the years around the way ... Because they just bought a lot of companies, the integration has always been hard, right? Because you have different systems that they bought many different companies. Slack is more notable. Obviously, Tableau and others, Heroku, and tons of other companies. I recently have come out and said, "Hey, I think they have an opportunity." Everyone is like, "Oh, Salesforce will get crushed by GenAI. We'll build Salesforce here." Maybe true, but they have an opportunity, because they have data.
Jake Graham
>> Yep.>> So Salesforce has a ton of data. So opportunity for Salesforce has abstract away. Now, they've got the data cloud, they're working on this. Their stock took a hit yesterday, but their cash flow is up, so they have a ton of cash. They could build an abstraction layer.
Jake Graham
>> Absolutely.>> And then take all that siloed and just make it invisible. This is the promise of what the things that you're working on. What's your opinion of that? Is it doable for Salesforce? Do they have all the right data sets? What would you do if you were at Salesforce?
Jake Graham
>> So instead of just Salesforce, can I frame it into the largest software business?>> Yeah, sure.
Jake Graham
>> So let's say Salesforce, SAP, Workday, ServiceNow, I honestly believe there is a right and a wrong answer here. I believe the wrong answer is trying to put up barriers and close your eyes and say, "We can push data gravity back. Don't let anyone get access to the data we have. That's our moat." I don't think that works. What I do think will work is leaning into the idea of interoperable sharing will allow your customers to not only get value from the data that's in your platform, but also make them willing to give you access to data that doesn't live in your platform. So if any one of these companies, I would be doing really two things. I would be offering creating data products that are interoperably shareable between your customer's existing data states and the data that lives in your platform. If you put up walls to get data out, those are also going to be walls to get data in. Nobody wants to put their data into a place that they don't do as fair.>> And agents want to have authority and delegation between other agents across platforms.
Jake Graham
>> Exactly. We're getting to a place in which rigid UIs and even rigid queries are not going to be how data is accessed. There's going to be constant arbitrary questions being asked by users. And the answer, it's not going to be enough to say, "Sorry, that data is not in the Salesforce data cloud." It needs to be able to reach out and understand data wherever it lives. So I think the second thing I would do after offering this interoperability and extensibility is they have the opportunity to really win based on the value of those agents. They own not just the data but the workflow. And really what an agent is it's the ability to bring data to the workflow in a much more flexible way. So by opening the gates and allowing for this flow, it would allow them to actually differentiate and win on the power of those agentic workflows.>> Yeah. And generative AI is, I love the word generative because it means it's generative. It's new, it's a new prompt. It's not a prefabricated prompt. I was talking to Jim Zemlin who runs the Linux Foundation. He and I were talking this past summer in Paris at an event, and we were talking about MCP, you mentioned that earlier. And A2A is also out there. They're going to create two projects for both, but MCP has been a nice organic, community-driven execution around this idea of kind of API-like thinking around, "Hey, my MCP will talk to your MCP for data transfer." But it's not that easy. It's easy to say. I mean, it's actually a good mechanism, but still some things to work on, but the idea is simple. It's not just a black box. It has to have, you got to know what's on the other side. It's kind of a data supply chain challenge and delegation authority. These things are kind of concepts. What's your view of MCP and A2A? How does that have to work? Because I think you're right. I think these walls can't have walled gardens. You got to be open. But the context now, to your point about context is interesting. If I have a data product that's available that could be shared, that could be consumed through an MCP mechanism.
Jake Graham
>> Exactly, and that's exactly what we're working on. I mean, one, I have been surprised and thrilled by how fast there's been cross-platform adoption of MCP. I think that's going to enable our industry to grow significantly faster than it would have otherwise, but I do think people tend to overestimate what MCP is. MCP is an agent-native way of describing functionalities of an endpoint. And I think that, that's fantastic, but it depends on whatever the capabilities of that endpoint, what they are. And I don't think we've got a strong understanding of where does the intelligence layer actually live. So if I take, for example, our customers. Today, many of them provide data as a service, but that actual intelligence often lives with the customer who is determining, "What model am I going to build? What questions am I going to ask?">> This is the context.
Jake Graham
>> Where I think that's going to shift is data as a service, the companies becoming insider knowledge as a service companies in which you pass language to an MCP endpoint and that language is actually interpreted and returned into data. And I think discovering where the insight and where that intelligence layer lives on a use case by use case basis is actually going to be a lot of where the turf wars exist. So if I were sitting, again, if I'm Salesforce or SAP or ServiceNow, I'd be less worried about whose database or data lake is this sitting in and more worried about where does that intelligence layer sit and how do you get the right access to the right data for it.>> Your premise is that intelligence layers should sit with the customer, or maybe not.
Jake Graham
>> My premise is I think everyone should be racing as quickly as possible to own as much of the intelligence layer as possible. I think that if you want to differentiate and you want to build real value in the world, you actually want that intelligence layer to live with you, whatever your organization is. So as much as possible, I want the ability to translate a question into a SQL query, an interpretation of knowledge to live in Bobsled for our customers. And I think there's going to be many of other cases in which it should live downstream, but that to me is where the turf war should be, not whose database is this data sitting in.>> So you're basically saying it's situational?
Jake Graham
>> Exactly.>> And if the human in the loop is leveraging data, they should have as much intelligence at their fingertips as possible.
Jake Graham
>> Exactly. And that what every technology business should be doing as quickly as possible is focusing on building that intelligence layer, not making sure that you can't use their API for whatever.>> It's kind of a compliance mindset, protective. It's like defense versus offense.
Jake Graham
>> Exactly.>> Play more offense.
Jake Graham
>> This is an offense moment to me, not a defense moment. I think that's a fantastic way to put it.>> Great conversation, but I didn't even get to half the questions I wanted to ask you. Okay. Let's kind of wrap up. Just give us a taste of what you guys are working on now. You guys are growing, how big is the company? What are your goals? What's your optimization plan?
Jake Graham
>> So we're at 30 people now. We've grown about 10x over the last 18 months. And where we're focused from a product perspective is really being the answer for where you create this context layer for data products. I think from a company growth perspective, we expect to come out to market and raise additional funding over the next year and really become the answer for how do you create data products that are consumed wherever your customers were.>> Well, I certainly endorse your approach. And if you're watching and you're an investor, you get 10% discount on the round.
Jake Graham
>> There you go.>> Just so you know, John recommended you. Jake, great venture, love what you're doing, offense not defense. Age is going to be there. It's going to be reality. And if you don't have the data connected and intelligently set up-
Jake Graham
>> I totally agree.... >> you're in the wrong side of history.
Jake Graham
>> And I appreciate the plug. We'll be sure to save you some allocation.>> All right. I'm John Furrier with theCUBE. We are here with a mixture of experts. And speaking of AI, check out thecubeai.com, that's where all the mixture of experts, brains will be set, query. We've got a light search engine there, check it out. And of course, we're doing our part to bring you all the best data here from the experts. Thanks for watching.
