About this role
As a Research Scientist on the Agent Flywheel team, you will build the core systems that enable recursive self-improvement for AI agents. You will transform raw customer interaction data into actionable signals that drive agent optimization and product development. This role offers a unique opportunity to advance the state-of-the-art in enterprise AI and directly impact products used by thousands of organizations.
What we are looking for 6
- Develop context layers for recursive self-improvement and autofix loops in AI agents
- Convert messy customer interaction traces into deterministic metrics and online signals
- Train and calibrate LLM judges for reliable model feedback
- Design benchmarks and RL environments from real customer workloads
- Advance research in online evaluation and automated measurement
- Apply expertise in debugging complex, stochastic systems