Full-stack engineer, LLM features
Own the layer between raw events and a plain-English answer. You'll build the query surface teams use daily, and the LLM features that turn “why did activation drop?” into a real, cited result — with guardrails so it never invents a number.
You'll be doing
- Natural-language querying over the event store
- Auto-detected insights: anomaly → explanation → link
- Product UI end to end — React, TypeScript, our design system
You probably have
- Shipped a product where you owned both ends
- Put an LLM feature in front of real users, evals and all
- Opinions about latency, and the receipts to back them