Sr. Research Scientist - Agent Flywheel
Core
Build the context layer for recursive self-improvement and autofix loops for agents by converting raw production session traces into online signals, calibrated LLM judges, and failure clusters.
Role type
Senior Research Scientist (Multi-Agent Systems & Evaluation)
Builds
Online signals, calibrated LLM judges, datasets, reward signals, and RL environments for agent improvement.
Domain
Enterprise AI, Multi-Agent Systems, Reinforcement Learning
Deliverable
production ML models
Required skills
Debugging complex stochastic systems, Designing metrics and evaluations, Experimental rigor and causal thinking, AI research, Software engineering
Preferred skills
Agent systems (coding, long-horizon tool-using), Learning from feedback (post-training, RLHF, reward modeling)
Technologies
LLMs, RL environments, Deep learning frameworks
Responsibilities
Convert multi-turn agent trajectories into datasets and reward signals for post-training. Design benchmarks and RL environments from real customer workloads. Advance evaluation research including online evals on live traffic and automated raters. Keep up with latest developments in deep learning to create new techniques.
Seniority
Senior, hands-on IC