(Kaggle, DrivenData, etc.)

You’ll draw on your competition experience, mathematical reasoning, and programming skills to create tasks that require more than a plausible answer. Each task should have a reproducible environment, a verified reference solution, and reliable checks for successful completion.

• Build reproducible environments, prepare relevant data and dependencies, and implement reference solutions.
• Develop tests and evaluation criteria that reliably distinguish correct solutions from incomplete or flawed approaches.
• Run coding agents against tasks, investigate their failures, and refine the tasks based on observed behavior.
• Identify implementation errors, flawed assumptions, data leakage, and other issues that affect correctness.
• Document task requirements, solution reasoning, and verification steps clearly.

• Demonstrated experience in data science, ML, quantitative modeling, or relevant research competitions. (Kaggle, DrivenData, etc.)
• Strong problem-solving ability and a clear understanding of the methods you use.
• Practical programming ability sufficient to implement and verify your own solutions.

Relevant competition platforms include Kaggle, DrivenData, Zindi, AIcrowd, Numerai, DataSource.ai, and CodaLab/Codabench.

Work arrangement: Fully remote

Eligible locations: Worldwide