Data meets AI
Data systems are evolving with AI. AI is not only changing how we build, query, search, and observe modern data platforms, but also bringing entirely new types of data and workloads into the picture.
Join us for a morning of databases and AI, as engineers from the community share practical experiences of bringing data and AI together.
From finding the right way to search through billions of vectors, to making databases conversational, to understanding what it takes to observe modern AI platforms - this meetup will explore the systems quietly powering the AI revolution.
Whether you're building AI applications, working on data infrastructure, or simply curious about what happens when Data Meets AI, this meetup is for you.
Don't miss out! RSVP and secure your spot!
Agenda
09:30 AM: Registration & networking
10:00 AM: Welcome & opening
10:15 AM: Talk 1 - Chat with Your Data: Beyond SQL
10:45 AM: Talk 2 - Vector Search: Do you really need another Database?
11:15 AM: Break
11:30 AM: Talk 3 - Optimizing Inference with ClickHouse-Backed Observability
12:00 PM: Talk 4 - The Model Gets the Headlines. The Database Wins the Race
12:30 PM: Q&A with Speakers
12:55 PM: Closing Remarks
01:00 PM: Lunch and Networking
What to expect
Expect deep technical talks and real-world engineering stories around the intersection of data infrastructure and AI.
Expect practical insights, architecture deep dives, and lessons from building and operating these systems in the real world.
We'll wrap up with networking over lunch, giving you a chance to exchange ideas and experiences with fellow data engineers, database enthusiasts, platform engineers, and AI builders.
Bring your questions, your experiences, and your curiosity about the systems behind AI.
***
Session Details: Chat with Your Data: Beyond SQL
Join us for a practical look into building a conversational natural-language interface for databases from the ground up! We’ll dive into how we integrated this system with ClickHouse to make complex data exploration effortless across different engineering and business roles.
Beyond the feature demos, we’ll share the real-world problems this approach solves, the core engineering challenges we ran into while building it, and key technical lessons learned. Whether you're curious about AI-driven database tools, LLM integrations, or analytics user experience, you’ll walk away with actionable insights for your own projects.
Speakers: Sachidananda Maharana, SMTS @ Nutanix & Punith Subashchandra, MTS-3 @ Nutanix
Session Details: Vector Search: Do you really need another Database?
Vector search has quickly become a key building block for AI applications—but does it always mean introducing a new database? This talk will take a journey through what vector search is, how it evolved, and where it actually adds value. We’ll explore real-world use cases that benefit from vector search, the trade-offs involved in choosing a vector database, and how to evaluate whether a specialized vector database—or an existing database like PostgreSQL or ClickHouse—is the right fit for your workload. We’ll look at what capabilities existing databases already provide for vector search, where they work well, and where a dedicated vector database can offer significant advantages.
Speakers: Pranav Mehta, MTS -4 @ Nutanix & Kailash Kejriwal, MTS-1 @ Nutanix
Session Details: Optimizing Inference with ClickHouse-Backed Observability
Inference performance is workload-specific. A high-throughput batch summarization job cares about total tokens processed per second and can tolerate a slow first token. A coding assistant optimizes for time-to-first-token instead, since a developer is watching the cursor. Coding assistants also reuse context across requests - the same system prompt, the same file. That makes prefix caching one of the biggest levers on cost and latency.
Agentic, multi-step tool-calling workloads look different. One request turns into a chain of calls: plan, call a tool, reinject the result, repeat. The chain length isn't known upfront. So the metric that matters isn't the speed of any single call - it's how long the whole task takes.
For a workload shaped like this, tuning performance means leveraging two views: an aggregate one to spot that task latency is regressing (Metrics), and a detailed one to find which step in which chain is responsible (Traces).
This talk is a case study of optimizing one such agentic workload on Nutanix Enterprise AI (NAI).
We used ClickHouse as a single backend for both metrics and traces, via Jaeger. That let us go from "something's wrong in aggregate" to "here's the exact task, and the exact step, that shows why", without switching systems!
We'll cover how Jaeger's parent-child span model maps onto an agent's call chain.
We'll walk through how we structured that data in ClickHouse, how it stayed fast even with many different tags to filter and group by, and how ClickHouse's query cache kept our repeated debugging queries cheap. Finally, we'll share what we found and fixed - and how we're now rolling out the same approach to other inference workloads
Speakers: Shiv Jha, Staff Engineer @ Nutanix & , Aishwarya Raimule, MTS-4 @ Nutanix
Session Details: The Model Gets the Headlines. The Database Wins the Race
Everyone is arguing about which model is best. Almost nobody is talking about the system underneath it. In F1, the driver gets the trophy, but the car decides who is on the podium. Production AI works the same way: the model is the part users see, while the data layer decides whether your product feels instant or unusable. This talk maps the F1 grid onto a production AI system: models as drivers, data as fuel, observability as telemetry. Then it shows the full loop live: natural language questions over F1 race data through LibreChat, an agent querying ClickHouse via MCP, and every request traced in Langfuse, with latency, tokens, and cost queryable as plain SQL. You'll leave knowing how to instrument your AI system like a race engineer, and why the teams that win are the ones that build the whole car.
Speakers: Siddhant Agarwal, Senior Developer Relations Advocate - APJ @ ClickHouse
See you at the meetup!