Competitive Data Scientist (Kaggle)

About the role

Suits a competitor who likes breaking problems into testable pieces and debugging other people's code. Work centres on tasks that need real reasoning, with reproducible setups and verified reference answers.

At a glance

What you will do

• Design data science, machine learning and modelling tasks drawn from real problems and contest experience
• Build reproducible environments with datasets, dependencies and consistent execution
• Write efficient reference solutions and automated tests that separate correct work from flawed work
• Run AI coding agents on the tasks, study failure patterns and tighten the specifications
• Find bugs, data leakage and weak methods, then document problems, outputs and checks

What they look for

• Hands-on data science, machine learning, quantitative modelling or applied maths background
• Competition experience on platforms such as Kaggle, DrivenData, Zindi, AIcrowd, Numerai or CodaLab
• Python skills good enough to implement, test and debug solutions alone
• Grasp of statistics, preprocessing, feature engineering and model evaluation
• Careful reading of results and clear written technical reasoning

A typical week

• Flexible schedule, including weekends, at roughly 15 hours a week
• Move between task design, environment setup and agent evaluation
• Iterate on task difficulty and reliability through repeated testing