Inside the Intelligent Data Lakehouse with Rahim Bhojani

Drowning in data, but starving for insights?

Rahim Bhojani, CTO at Dremio, breaks down how intelligent data lakehouses and self-service analytics are helping organizations cut through the noise. From enabling faster, AI-powered insights to reducing dependency on data teams, he shares how businesses can make data truly accessible, actionable, and impactful—at scale.

Having spent over 20 years in data infrastructure and self-service analytics, what key shifts have you observed in the data ecosystem? How do they influence your approach as Dremio’s CTO?

Over the past two decades, the data ecosystem has shifted dramatically. The most noticeable change has been the rise in data and tech fluency across organizations. Access to data is no longer limited to specialists, as business users, analysts, and developers now expect self-service capabilities as the norm.

Cloud was the biggest productivity booster in this era. What once took months—provisioning hardware, managing capacity—can now be done in minutes. The ability to instantly scale and experiment has removed barriers to innovation and accelerated the pace of development across the stack.

Now, AI is driving the next wave, which is further reducing friction and expanding access. At Dremio, we see this as a continuation of our mission: democratizing data access and delivering performance at scale. AI agents are simply new clients—issuing queries, interpreting results, and driving decisions. Our platform is evolving to serve them just as it does analysts and applications—securely, efficiently, and at scale.

Many platforms promise performance and scalability; how does Dremio deliver on both, without sacrificing cost efficiency or flexibility?

Many platforms claim performance and scalability, but often at the expense of cost or flexibility. At Dremio, we approach this differently by architecting for efficiency from the ground up.

Our foundation is the data lakehouse, which inherently supports scale by decoupling storage and compute. This separation allows organizations to store massive volumes of data cost-effectively in open formats like Apache Iceberg, while elastically scaling compute resources based on workload demands.

On top of this, Dremio delivers high-performance SQL through a cloud-native engine that pushes processing down to where the data lives and minimizes movement. We further accelerate queries using Reflections—optimized, relational caches of source data that enable our query engine to partially or fully satisfy requests without hitting the underlying storage. These are created autonomously based on query patterns and support aggregations, raw data, and join-optimized layouts, delivering fast results without manual tuning.

And with Dremio deployed in your cloud or datacenter, using your storage and security controls, you get full flexibility without lock-in—combining warehouse-level performance with lakehouse economics. Additional infrastructure is only needed as your use cases and query volume grow, not just to maintain baseline performance.

In short, Dremio delivers performance and scale through architectural intelligence rather than brute force, so you only pay for what you need, when you need it.

Take us through the latest Dremio MCP Server. What agentic AI capabilities set it apart from similar offerings?

The modern data stack is evolving rapidly around the agent experience. With the viral adoption of tools like Claude and ChatGPT, natural language is quickly becoming the standard interface for analytics and decision-making workflows.

For Dremio, this is a natural evolution, not a reinvention. Thanks to our semantic layer, our platform has always been about providing a single pane of glass over data spread across the enterprise. Historically, the clients were tools like Tableau, Power BI, and Jupyter notebooks. Now, AI agents are emerging as the next generation of clients. And these agents will add to data sprawl and require the same foundational things: fast, secure, and governed access to data without having to copy or move it.

With our implementation of the open Model Context Protocol (MCP), Dremio integrates natively with leading LLMs like Claude and ChatGPT,  giving customers a standard way to build their own agents and AI apps. Agents can dynamically discover and invoke platform capabilities, like running SQL, accessing lineage, or exploring metadata, creating data models just as a human user would, but programmatically and at scale. Dremio is not just compatible with AI—we’re evolving with it. We’re making sure that agents, like users before them, can access enterprise data safely, efficiently, and intelligently.

Open standards such as Apache Arrow and Iceberg are core to Dremio. How do you shape engineering strategy around open source while still building differentiated value?

Open standards and open source are foundational to how we think about engineering at Dremio. Technologies like Apache Arrow, Iceberg, and Polaris (incubating) are not just implementation choices; they reflect a core belief: customers deserve choice.

Our mission is to meet customers where they are, whether they want to run on their own cloud, use open formats, or integrate with a broader ecosystem. Building on open source allows us to avoid lock-in, interoperate natively, and empower customers to design the architecture that works best for them.

From a strategy perspective, we invest heavily in open source innovation: Arrow for in-memory performance, Iceberg for transactional tables at scale, and Polaris (incubating) as an open catalog with cross-platform interoperability. These aren’t just enablers for our platform; they’re enabling a broader open data ecosystem.

At the same time, we focus on delivering differentiated value on top. That’s where Dremio brings unique capabilities such as our semantic layer, Reflections for autonomous acceleration, multi-engine isolation, and now AI-native features through the MCP Server.

By standing on open foundations and building differentiated, opinionated experiences above them, we ensure customers have both freedom of choice and the performance and simplicity they expect from a modern data platform.

What does long-term agility look like in a lakehouse platform, and how do you design for it without adding complexity?

This is a great question because while the lakehouse has become a popular concept, it’s often misunderstood as a single product. In reality, it’s an architectural approach made up of many components: open table formats like Iceberg, cloud object storage, multiple query engines, catalogs, semantic layers, and now increasingly, AI agents. The promise is openness and flexibility, but it also introduces complexity. Customers often face fragmented metadata, inconsistent performance, governance sprawl, and operational overhead. And with the rise of AI, traditional platforms aren’t built to support agentic workflows natively.

Another major challenge is that customers own their storage in the lakehouse model, which means they also own the physical layout of the data. If that layout isn’t managed well, it leads to degraded performance and unnecessary costs. At Dremio, we’re solving this by abstracting away physical optimization through features like Iceberg Clustering, automated table maintenance, and intelligent caching via Reflections—so users get high performance without manual tuning.

Long-term agility in the lakehouse era means giving customers openness and choice—without the complexity. That’s exactly what we’re focused on delivering.

How do you see the roles of data engineers, analysts, and product teams shifting in response to more democratized access to data?

This is a pivotal moment for data teams, but it’s also part of a longer arc. Self-service has been growing for the past 10–15 years, steadily raising the level of data literacy across organizations. Now more than ever, people—from analysts to product managers—are fluent in data. AI is accelerating that trend. It’s the great equalizer, giving non-technical users the ability to perform complex data tasks through natural language. As a result, roles are evolving, and in many cases, becoming more interchangeable. Analysts are moving beyond dashboards to drive strategy. Product teams are querying data directly. And data engineers are shifting from pipeline builders to platform architects, focused on scalability and governance.

In this new landscape, the lines between who asks the question, who prepares the data, and who interprets the result are increasingly blurred. What matters most is giving the right people the ability to act on data, regardless of title.

What do you see as the next frontier for “intelligent” data platforms, beyond performance and storage optimization?

The next frontier for intelligent data platforms goes beyond performance and storage optimization, which are now table stakes. What’s emerging is a focus on accessibility and usability at scale.

This is because acquiring data is becoming dramatically easier. With the explosion of APIs, streaming sources, and third-party connectors, data is more readily available than ever. At the same time, consumption is being democratized, meaning anyone, regardless of technical skill, can interact with data through natural language, no-code tools, or AI agents.

This shift puts a spotlight on the semantic layer. It has to do more than just abstract complexity; it needs to serve both humans and AI. That means translating business context into something that can be interpreted by a dashboard, a prompt to a language model, or a Python script. And it needs to be omnichannel so it is accessible through SQL, GUIs, notebooks, or conversation.

In short, the future isn’t just about faster queries or cheaper storage; it’s about intelligent interfaces that empower anyone to use data, with trust, consistency, and context built in.