You will turn competitive data science problems into clear, reproducible tasks for AI model development. The work combines programming, mathematical reasoning, software testing, and careful review of coding-agent behavior.
You will build environments with the required data, dependencies, and reference implementations. You will also run coding agents, investigate failures, and improve task design based on what you observe.
- Build reproducible environments with data, dependencies, and reference solutions.
- Create tests and evaluation rules that identify incomplete, incorrect, or unreliable implementations.
- Review coding-agent outputs and investigate why solutions fail.
- Find implementation errors, flawed assumptions, data leakage, and other issues that affect correctness.
- Document task requirements, solution reasoning, and verification procedures clearly.