Cloudera EVOLVE 2025 | Jason Mills, Cloudera

Clips
More from Cloudera EVOLVE 2025

Jason Mills

SVP Solutions Engineering

Cloudera

play_circle_outline Jason Mills' career highlights, including roles at Citi, J.P. Morgan, Google, and Snowflake.

play_circle_outline Cloudera's Vision: Unifying Open Source for AI Success Through Strategic Data Management and Future Growth

play_circle_outline Overcoming 'Pilotitis': Key Strategies for Effective and Scalable AI Deployment Through Accuracy, Cost-Effectiveness, and Security Governance

play_circle_outline The significance of the Tycoon acquisition for Cloudera's multi-cloud and on-prem deployment capabilities.

Info
Transcript

Jason Mills, Cloudera

Jason Mills

SVP Solutions Engineering Cloudera

In this exclusive interview for Cloudera EVOLVE25, theCUBE analyst Dave Vellante speaks with Jason Mills, senior vice president of solutions engineering at Cloudera, about the company’s push to redefine enterprise AI. Mills draws on his experience at Citi, J.P. Morgan, Google Cloud and Snowflake to highlight why a strong data foundation is essential to any AI strategy. Cloudera is simplifying open-source capabilities into one governed platform that supports on-premises, cloud and hybrid environments, he explains.

The conversation unpacks economics, i... Read more

explore Keep Exploring

What has been the career trajectory and experiences of the individual discussed in the text? add

What role does Cloudera play in assisting organizations with their data strategies and how do they adapt to the diverse data environments of their customers? add

What factors differentiate successful deployment of AI models in production from those that remain in pilot mode? add

What advantages does the Kubernetes engine Tycoon provide to Cloudera as a platform? add

bolt Powered by CUBE AI

Jason Mills, Cloudera

search

Dave Vellante

>> Welcome back to New York City, everybody. We're live here at Cloudera Evolve. My name is Dave Vellante. We're here in Hell's Kitchen. Yeah, kind of a murky day, but beautiful view of the river. I'm really excited to have Jason Mills here as the senior Vice President of Solutions Engineering at Cloudera. Jason, thanks for coming on. Appreciate your time.

Jason Mills

>> Thanks for having me.

Dave Vellante

>> You're very welcome. You've got an amazing background that I'd like you to talk about before you talk about your role here, but you've been at big banks, you've been at Google, you've been at Snowflake. Talk about your career and how you got here and what you do here.

Jason Mills

>> So first of all, thank you for having me. And yeah, I've been blessed to have an amazing career going into organizations at major inflection points. And so I spent a large part of my career, 20 plus years in financial service technology, 10 years at Citi, 10 years at J. P. Morgan. But the last five years at J. P. Morgan around the 2015 era, we looked at the ability of creating an AI center of excellence, and I helped run that. And ironically, the amazing thing about that was it was all run on Cloudera technology. This was the early days of Hadoop. And the idea was figuring out how we could take that data and either commercialize it or use it for the bank's benefit. And so that was an amazing time. Learned a lot, did a lot of amazing use cases. Fast-forward, landed a great role at Google Cloud at its turning point, right? 2018 or so when the company was really trying to redefine its place in the cloud market. Amazing four years there, worked with some amazing people, did some amazing things with the largest customers in the world. Then landed a great role at Snowflake, leading customer engineering, solution engineering, and spent three and a half years there redefining what it meant to sell or talk to customers about cloud native data warehousing. And then most recently this year, I got a call from Cloudera and they said, "Listen, we'd love for you to do what you did there in your previous roles here at Cloudera and help us along this journey of redefining Cloudera's position in the market." And I'm super excited to be here. We're doing some amazing things. I've been here four months, I've traveled to four countries and have had terrific conversations with customers about what they're doing with our products.

Dave Vellante

>> So now is the time for you to help Cloudera do that? I think a lot of people may not understand or may not remember. So when Hadoop ascended, it was non-trivial to get, I mean, it was easy to cut your storage costs. That was kind of the no-brainer part, but actually extracting value. You needed a lot of skills. And so those big banks had those skills, the big internet companies had those skills. And so Cloudera has a big footprint in all those organizations and now they're adding value on top of that. So you serve some of the biggest organizations in the world. Where do you see them headed? What are they adding to that original big data infrastructure, and how do you see Cloudera helping them get there?

Jason Mills

>> Well, the name of the game today is all about understanding where and what to do with your data. And so as we think about AI, machine learning and a lot of use cases that create automation for our customers, the first question they're asking is, what is our data story? And so if you think about what Cloudera does as a company, as a platform, takes highly complex, generally open source capabilities, brings them, simplifies them into a single platform and offers them as value to our customers. And so in that sense, customers are looking to us to help them define their data strategy. You don't have an AI strategy without a data strategy. And so the first point is defining that. And now what we're offering is a hybrid approach to that because we know our customers are not monolithic in their data use. It's data location on-prem, cloud, private cloud. We're basically saying to customers, "We can help you manage that data wherever it sits. And further, when you want to turn that data into highly used automation, you can unleash capabilities that are secure, governed and scalable on any environment under our platform."

Dave Vellante

>> There were a lot of good data charts in the keynote this morning. I took a lot of pictures. One that really stood out to me was Abbas, who's the chief strategy officer. He had a chart that showed the economics of adding workloads in the cloud and instantiation. And I think it was at a hundred workloads, you start to get essentially negative economies of scale and the cloud and you get operating leverage and using the Cloudera model. And so I'm interested in how clients are getting sort of that repeatable business value. It seems to me if you've, you've got diseconomies of scale, it's going to be harder for you to add on workloads and get that repeatable value. At scale, it sounds like you've got some advantages with regard to helping customers get that repeatable business value. Now, I know it's early days, most customers, they're still struggling to get ROI, hear about that, but it's only a matter of time before the operational model and the business model kicks in. So I'd like you to speak to that.

Jason Mills

>> Yeah, so what we're talking about is the difference between on-prem and cloud deployments of machine learning or AI models. And what we've seen for those customers who own their own data centers, they're able to differentiate between either experimentation or full production inferencing on-prem and make a decision with the cloud, our platform, whether or not to move that workload to the cloud. And for some of our largest customers, we're even looking at ways to burst into the cloud so that they can have hyper-agile solutions within their environment. So cloud costs and on-prem costs are a little bit, they're not apples-to-apples measurements. And so when you think about the AI ecosystem, there's a lot of sub-components that define what that cost is. At the forefront, you've got the early stage, the data preparation, the way that you get data ready for an AI. That could be labeling data, that could be data preparation, even data movement all has its early on costs. The second phase of that is looking at the capabilities that you want to leverage. Which models do you want to use open source models, do you want to have your own models? Do you want to retrain those models? You want to use RAG or retrieval augmented learning on those capabilities. And then that decision is done across many weeks or months, even quarters of experimentation.

Dave Vellante

>> And the models change as -

Jason Mills

>> As you're doing those POCs. And so what our customers are looking to us for is a couple things. Number one, interoperability, with a variety of data sets that exist either on-prem or in the cloud that may be locked in silos. So we offer that interoperability. The second thing they're looking for from us is simplicity in using the platform. And we are improving our capabilities around user experience, user interface, so that it's less clicks to get to the same destination. That's our goal. And so providing that to not only deeply technical users like data scientists and data engineers, but simplifying the platform so that even business analysts can begin to ask questions of that data and ultimately show the scalability of the platform. And then lastly, when you think about costs, you also have to think about where you want to run that compute. Are you running that compute in the cloud where you have a significant scalability on GPUs and CPUs environments? Or do you may want to run that infrastructure on-prem where you have more controlled costs? Right? And so we're helping customers think about the holistic picture of managing ai, whether it's on-prem or in the cloud, and we're helping our customers or make the right decision for the right use case.

Dave Vellante

>> Quick side question on what you just said around business analysts. A lot of people think, "Oh, well, business analysts, they're going to be out of a job because AI's going to take that job." And people, Jensen, and many of others have said it, "No, it's going to be somebody using AI that ultimately takes that job." So two things. One is you talked about reducing clicks, we heard that this morning. How viable is it that we're going to be speaking to the system? And what do you see happening to that business analyst? Many of the business analysts I talked to are like, "Wow, I'm becoming actually a decent Python coder and I really wasn't a good coder before and I'm just elevated. I'm so much smarter than I was before." So sort of a two-part question there. Will we be speaking to the system and will those business analysts that gain superpowers so that they're actually elevated, not deprecated in the model?

Jason Mills

>> So there's a variety of ways to interact with your AI models or your data. And natural language processing is at the heart of what we use the capability techniques that we use to allow humans to speak to their data. And so this natural language querying of your data is at the forefront, and we're seeing less technical users want that capability, almost like a chatbot for your data. Second, we do have other communities of technical users, your Python developers, your SQL developers, all needing more diverse access to data sets that may be siloed in different locations. And so what they're doing is being able to pull that data into one location or believe just access that data from a variety of locations and provide that value to either external customers, partners, or even their internal users. And so our goal is to create a system that supports that entire community, that ecosystem of business users. Some will be augmented with their ability to ask questions of the data. Some will be drivers of new use cases, but because they're using Python or SQL to create new applications to enable customers to interact with that data. a platform in the middle of all that. That's what Cloudera is. It represents that platform that you can use work on-prem or in the cloud to provide all of those user communities with access intelligence to their data.

Dave Vellante

>> Brian in this keynote, Brian from ABVI, he had an interesting word that caught my attention. He said, the biggest mistake of scaling is you let tech, the shiny new toy, drive the deployment and that sticks people in pilotitis. Pilotitis is a word I hadn't heard before. It was a very good term. So I'm wondering what patterns do you see that's going to separate organizations from pilotitis so that they can get to scale?

Jason Mills

>> So I think one of the things that differentiates me in my career is I spent 20 years on the buy side. I was a buyer of these systems and platforms, and now I'm on the sell side where I'm selling and communicating how we can provide value to our customers. And one of the things that I constantly pay attention to is the ability to take AI workloads or any production, any system into production. What does production actually mean? Production means that you have a variety of parameters that you need to make sure or operating efficiently and cost within your cost range. And so if an AI model is not highly accurate to the questions or use cases that you're trying to address, you will stay in pilot mode. If that AI model costs more than you actually forecasted, and there's various ways to do that, then again, you will stay in pilot mode. And lastly, if it doesn't provide the security and governance over the entire ecosystem so that you can trust the output of those models, you will also stay in pilot mode. And so that's part of the understanding how customers need to pay attention to and rely on products like platforms like Cloudera to provide them with those capabilities to govern the entire AI ecosystem, if you will. And I'm seeing more and more customers trust the Cloudera platform for that, specifically our AI studio capability and our workbench capability to deploy these pilots, these science projects into production and seeing real upload and value for that.

Dave Vellante

>> I want to ask you about use cases. Jensen lays out at GCT, three vectors of growth for AI, AI in the cloud, AI and enterprise and AI and robotics. I'm interested in the middle because that's really where hybrid AI and private AI start to come into play. And I've said that most enterprises today, they're doing their experimentation in the cloud or there's training going on in the cloud. That's a really exciting use case. We all know about that. But most enterprises are playing, I call it small ball, and they're hitting singles and they're maybe taking walks and working the count as a use of baseball analogies, but that's okay. That's where they're going to get good returns. There's maybe some boring use cases, but where are you seeing some of those quick wins?

Jason Mills

>> Yeah, so the quick wins, I don't know if we call it quick wins because a lot of this has been the 80/20 rule. You have 80% of the time you're spending on data collection, data cleansing, data preparation. So there's a lot of work that's been done over the years to get to a point where we are today for the mature organizations. But what we're seeing wins are in what I call enterprise risk management. That is the best balance between, in some use cases, external ROI where we're seeing use cases like fraud detection, reducing overall company impact to fraud for our customers or internal improvements such as doc AI, which automates internal documents to improve processes like mortgage filing applications and things of this nature. So there are a variety of what we call enterprise risk management use cases that directly address the company's need to improve automation, augment human performance, as well as deliver ROI for the company.

Dave Vellante

>> So I want to ask you about the acquisition of Tycoon. We'll be at KubeCon next month.

Jason Mills

>> Yep.

Dave Vellante

>> Next month. Yeah, we'll be at KubeCon. We've been to every, KubeCon started with DockerCon. I mean theCUBE has been there. What did that acquisition do for you? It gives you mobility, but specifically certainly helped your positioning in the market, but also expanded your TAM. I wonder if you could talk about how you see that affecting the way in which organizations are deploying things like multi-cloud or across the states.

Jason Mills

>> So it's a great question in terms of what a Kubernetes engine like Tycoon delivers to a Cloudera as a platform. So prior to Tycoon, Cloudera use a variety of different ways to deploy the platform to our customer, both on-prem and in the cloud. So first off, Tycoon technically gives us a single Kubernetes engine across all of our deployment locations, whether it's a private cloud, on-prem or in the cloud, a single way to deploy that. So one code deployed to many locations. That's the first advantage. The second advantage is within the product itself. It gives us interoperability with not only siloed data sets or other environments, but it also gives us the ability to bring in new technologies like vector databases and include that into the ecosystem. All governed within our SDX, our security governance framework, to allow customers to use more capabilities within the Cloudera ecosystem. And then lastly, it's ease of use. When you think about the ability to use AI workloads with GPUs, CPUs, and other high performance computing environments, this Tycoon platform will give us the ability to manage those environments seamlessly, even on a reservation basis. And so where we didn't have those capabilities before, specifically on-prem, we now have those, and now that creates a synergy between cloud and cloud-like on the prem. And that's what customers have been asking us for many years.

Dave Vellante

>> As a practitioner and a technologist, you've seen a lot of waves and you've seen a lot of abstraction layers. VMware, Kubernetes, you were at Google. I mean Google created-

Jason Mills

>> Inventors of Kubernetes.

Dave Vellante

>> Yeah, exactly. And so we're always adding abstraction layers in tech because tech gets more complicated and we need greater simplicity. I mean, LLMs are a form of abstraction layer. Do you see a new abstraction layer emerging for AI or will Kubernetes, do you think, knowing what you know about it, how it's been applied at these companies that you've worked at? And as a practitioner, will Kubernetes be that abstraction layer for AI?

Jason Mills

>> I think of the technology, it is an abstraction layer for ease of mind, right? Customers want to make sure that we are deploying highly sensitive data and applications that use that data in secure environments, and Kubernetes delivers that for the enterprise. That's number one. But in terms of the overall AI world, I think the real differentiation is going to be what you mentioned before, providing a semantic communication layer into data for the end user, no matter who that end user is. And I think we're heading more and more towards that. LLMs have certainly exploded the opportunity up in terms of looking at how you can communicate and ask questions, prompt engineering, et cetera. We're going to see more of that. I think we're going to see more of this natural language communication with any kind of data set that you imagine, whether it's creating a movie or music or even querying your data behind the scenes. I think that is going to be the next kind of explosion in this abstraction layer. Because it doesn't matter your technical ability, it doesn't matter your technical depth, it doesn't even matter your creative ability, you can now become, if you will, an AI engineer, quasi-AI engineer by being able to communicate effectively with the platform.

Dave Vellante

>> And as long as that data is harmonized, which is a semantic layer does, then you'll be able to achieve that vision and serve agents. And Jason, tremendous having you on here, you're great guest. Really appreciate your time.

Jason Mills

>> Thank you for having me.

Dave Vellante

>> Thank you. All right, and thank you for watching, this is Dave Vellante for theCUBE team at Cloudera Evolve, we're live. Right back after a short break from New York.