Applied Researcher – Agentic RL

This is a fully remote position, open to applicants in California, +1 more state.

📋 Description

• Engage in reinforcement learning for agents that can reason over extended timeframes, utilize tools, collaborate with other agents, recover from failures, and learn from experiences.

• Design and implement RL and ML systems across a variety of algorithms, environments, harnesses, reward and credit-assignment strategies, evaluation methods, and scalable experimental frameworks.

• Collaborate with application teams, partners, and open-source contributors on adaptable methods and systems.

• Transform workflows into tasks, environments, evaluations, and reward signals.

• Conduct post-training experiments against real benchmarks to assess whether improvements arise from models, data, harnesses, inference, or infrastructure.

• Develop agentic judges and credit-assignment techniques for comprehensive trajectories and tool utilization.

• Ensure multi-agent and long-horizon environments can be trained reliably at scale while maintaining debuggable traces and comparable benchmarks.

• Identify reward hacking, harness behavior, data leakage, and infrastructure artifacts, followed by redesigning experimental loops.

• Take ownership of reusable environments, harnesses, rewards, evaluations, experiments, open-source systems, and research papers.

⛳️ Requirements

• Proven experience in hands-on RL training, scalable ML systems, agent environments, evaluation, inference/training infrastructure, or interdisciplinary research involving methods and systems.

• Practical experience in constructing and troubleshooting model-training, post-training, agent evaluations, or closely related experimental loops.

• Strong experimental judgment, including the ability to establish meaningful baselines, conduct thorough trace inspections, and provide clear failure analysis.

• Proficient in Python programming.

• Ability to create artifacts that other researchers and engineers can run, analyze, and build upon.

• Comfortable tackling ambiguous problems and ensuring that partners and teammates remain informed.

• Publications, open-source projects, benchmarks, or delivered products that showcase outstanding work are highly regarded.

🏝️ Benefits

• Equity incentives.

• Flexible work arrangements (San Francisco, New York, London, or remote).

• Visa sponsorship and relocation assistance.

• Professional development budget.

• Team off-sites and conference participation.