Turn your Kaggle or DrivenData competition experience into flexible remote work, about 15 hours a week, from anywhere. You design hard data science problems with reproducible environments, verified reference solutions and checks that separate correct work from flawed approaches.
What you will do
- Build reproducible environments, prepare the data and dependencies, and write reference solutions.
- Develop tests and evaluation criteria that reliably tell correct solutions from incomplete or flawed ones.
- Run coding agents against your tasks, study where they fail, and refine the tasks.
- Catch implementation errors, flawed assumptions, data leakage and other correctness issues.
- Document task requirements, solution reasoning and verification steps clearly.
What you need
- Proven experience in data science, ML, quantitative modelling or research competitions, for example Kaggle, DrivenData, Zindi, AIcrowd, Numerai or CodaLab/Codabench.
- Strong problem-solving skills and a clear understanding of the methods you use.
- Enough programming skill to implement and verify your own solutions.
- A big plus: strong Python, including testing, debugging and working in unfamiliar code.