RL Environments Engineer (AI Training Data Startup)

A post-Series A startup that supplies datasets and eval infrastructure to frontier AI labs is hiring several engineers to design datasets, rubrics and reward signals. You would collaborate directly with researchers at leading labs, go from a hunch to a live experiment fast, and see your output flow into large training runs.

Responsibilities

• Craft data slices and explore data formats that reveal where models fail in areas such as coding, finance and business-process tasks
• Create, and iterate on, grading criteria and reward functions used in RLHF and RLVR training runs
• Model how annotators behave and run experiments that strengthen specific model skills
• Develop quantitative methods for judging how good, how varied and how influential a dataset is when it comes to capability and alignment
• Manage pipelines built from both real-world and synthetic data
• Turn lab researchers’ training goals into concrete specifications for data and scoring

Profile

• One to eight years as a machine learning or software engineer
• Direct experience building reinforcement learning environments, or sharp instincts for designing agent tasks as an RL or ML researcher
• Strong full-stack or backend depth in Python, TypeScript or similar
• Signals of excellence such as time at an RL company, ML work at a top venture-backed startup, or published AI research
• A computer science degree from a leading program
• Bias for action and willingness to do tedious, difficult work

On-site in the San Francisco Bay Area. Visa transfers considered.