Research Engineer, Agent Reliability

About the role

The Reasoning Group treats reliability as the research problem. You will build the infrastructure that measures drift, verifies plans, and lets agents recover from their own mistakes.

What you'll do

• 01Build evaluation harnesses for long-horizon agent behaviour.
• 02Implement self-verification and recovery mechanisms.
• 03Instrument agents so their reasoning can be audited and replayed.
• 04Work with researchers to close the loop between evaluation and training.

What we look for

• ◇Strong software engineering and systems fundamentals.
• ◇Experience with language-model agents or reinforcement learning.
• ◇An instinct for what makes a system trustworthy in production.
• ◇Care for correctness under long, compounding sequences.