Lead RL Environment Engineer

Overview

Lead the design and delivery of RL environments built from supplied code and domain data for an applied AI research lab that produces training data and evaluations for frontier AI labs.

What you'll do

• Decide which data can become a useful environment and clarify what data a lab needs.
• Work backwards from existing benchmarks, or design suitable evaluations.
• Build rewards, graders and verifiers that score agent runs and resist shortcuts and rewards hacking.
• Run agent experiments and measure whether post-training improves performance.
• Ship containerised, reproducible setups in Docker, in a standard format such as Harbor.

What you'll need

• You have personally built at least one RL environment, with a runnable repo or inspectable artefact and an explanation of what you built.
• Strong ML and RL fundamentals, including gradient descent and how RL environments work.
• Strong recent Python.
• Linux, Docker, deployment and sandboxes.
• Verifier design that resists gaming.

Nice to have

• Training or post-training models.

Details

• Location: Mumbai, India (onsite).