Sr. Research Scientist – Agent Flywheel

P-1213

Company Description

At Databricks, we are obsessed with enabling data teams to solve the world’s toughest problems, from security threat detection to cancer drug development. We do this by building and running the world’s best data and AI platform so our customers can focus on the high-value challenges that are central to their own missions.

The Databricks AI Research organization is pushing the frontier of next-generation enterprise AI. We believe a company’s data is its greatest competitive advantage, and we’re building the models and agents that unlock it. Our work spans the full stack, from model training to advanced multi-agent systems.

Job Description

As a Research Scientist on the Agent Flywheel team, you will build the context layer for recursive self-improvement and autofix loops for agents: turning raw production session traces from Databricks’ data agents into online signals agents can optimize against. You will research how to convert messy, real customer interactions into deterministic metrics, calibrated LLM judges, and failure clusters, and close the loop so those signals drive training and agent improvement. 

You will join a collaborative team of researchers with diverse backgrounds and encode your findings into products our customers rely on. 

The impact you will have

• Turn raw production session traces from Genie Code and Genie One into online signals agents can optimize against.
• Train and calibrate LLM judges so model feedback becomes reliable, repeatable signal, including calibration against expert human ratings.
• Convert multi-turn agent trajectories into datasets and reward signals for post-training, closing the loop between production signals, evaluation, and agent improvement.
• Design benchmarks and RL environments from real customer workloads that probe long-horizon, tool-using agent behavior.
• Advance evaluation research: online evals on live traffic, automated raters, verifiable rewards, and measurement that keeps pace with agents that act.
• Keep up with the latest developments in deep learning and create new techniques that go beyond the state of the art.

What we look for

• Expertise in debugging and diagnosing failures in complex, stochastic, multi-component systems
• Deep instinct for designing metrics and evals that track what actually matters, and questioning noisy, sparse, or biased signals
• Experimental rigor and causal thinking, from offline ablations to online experiments
• A drive to turn measurement into improvement: closing the loop from signals to changes that move the numbers
• Solid AI research and software engineering skills
• Strong communication skills and excitement to work on a small, fast-paced team
• A drive to get your work into the hands of real users

Nice to have

• Experience with agent systems, such as coding agents, long-horizon tool-using agents, or building evaluation environments
• A track record of novel research on learning from feedback, such as post-training, RLHF, or reward modeling