About the job
As a Staff Software Engineer - Agent Quality, you will be a founding member of a new team focused on evaluating and continuously improving Databricks' AI Agents. You will design and scale the infrastructure, tooling, and developer workflows that let researchers and engineers evaluate agents rigorously — driving a flywheel where evaluation results feed directly back into agent improvement across the full lifecycle, from development and training to production.
Responsibilities
Stand up the foundational evaluation infrastructure for Genie Agents, enabling rigorous benchmarking, regression detection, and quality measurement across research and product teams.
Build the flywheel that connects evaluation results back into agent improvement — closing the loop between production signals, training, and iterative development.
Shape the long-term technical direction for agent quality infrastructure, with real influence over how Databricks measures and improves its first-party agents and agent development platform.
Help shape the long-term technical direction for agent quality infrastructure as Databricks expands its first-party agents and agent development platform.
Qualifications
Minimum
6+ years industry experience building software systems
Strong Python programming skills, with experience building production or research infrastructure
Experience building or operating distributed systems, data pipelines, or large-scale infrastructure with a focus on reliability, correctness, and operational maturity
Ability to design pragmatic but rigorous systems that produce trustworthy, reproducible signals for complex applications
Comfort working across ambiguous research and product boundaries, and partnering with both researchers and engineers to turn ideas into robust internal platforms
A high bar for technical quality, strong ownership, and the ability to influence roadmap and execution across multiple teams
Preferred
Experience with devtools, CI/CD platforms, testing frameworks, observability tooling, or benchmarking infrastructure
Familiarity with how LLM or agent quality is measured — whether through evals, experimentation platforms, or production monitoring