AI Product Engineer - ClickStack

About this role

Join us in building the AI layer for Observability!

ClickStack is the open-source observability platform we're building at ClickHouse — logs, metrics, traces, and session replays unified so engineers can find root causes quickly. The interesting work now is in the agent layer: systems that can investigate an incident at 2 AM, propose a root cause, and hand the on-call a concise summary by the time they've logged in.

We're hiring an AI Product Engineer to build agentic capabilities on top of a petabyte-scale observability platform, with a focus on developer experience. If you've been building agents, designing skills, and wiring up MCP servers — and you've hit the limits of generic copilots for production work — we'd like to talk.

What you'll do

• Build agents that investigate incidents. They surface anomalies, answer "why is production broken?", and use ClickStack as their substrate.

• Write skills, not just prompts. Build a library of reusable skills that captures how our team debugs, finds root causes, writes ClickHouse queries, and runs incident response, so agents pick up the right playbook instead of starting from scratch.

• Own the agent stack end-to-end. Context engineering, tool design, evals, tracing, cost. You're responsible for whether the agent works in production.

• Make ClickStack a great place to run AI workloads. Build the MCP servers, SDKs, and integrations that let customers' agents read telemetry, take action, and stay observable themselves.

• Work in the open. Collaborate with OSS contributors and customers, debug their problems alongside them, and feed what you learn back into the product.

• Tackle the hard parts. Latency, cost, context window limits, eval coverage, hallucinations on real telemetry.

Who you are

• You've been building agents long enough to have opinions — about context engineering, tool design, when to use a skill vs. a tool, what evals catch and miss, and where popular frameworks break down.