Real take-home assignments. Real data. Your own AI agent.

How it works

1. Pick a take-home

   Read the brief and the data description, exactly as a candidate would receive them.

2. Connect your AI agent

   Add our MCP server to Claude, ChatGPT, Cursor or Claude Code and sign in with your Lean Data Scientist account. No API key needed.

3. Analyse in the sandbox

   Your agent runs read-only SQL against the dataset on our side. You steer the analysis, check the reasoning and write up your answer.

The datasets stay on our servers — there is nothing to download. Query results are capped and logged, which keeps the data available for everyone who practises here.

The take-home library

14 assignments from real hiring processes, covering A/B testing, predictive modelling, marketing analytics, operations and more.

Dataset:ab_testing

Evaluate a push-notification A/B test for reviving dormant users

The dataset is from a social app for mobile gamers that enables users to watch and interact with live streams and recorded videos of mobile…

Read the question →

Dataset:social_media

Estimate the revenue lift of an Android-only launch and project it to iPhone

A social media app is launching a change to its 'Trending' product. Because of the Google Play store release schedule, they have to launch…

Read the question →

Dataset:grocery_shopping

Predict how long a grocery shopper takes to pick an order

Welcome! We are a grocery delivery company that runs on data - and our awesome shoppers. One of the important ways we make customers happy…

Read the question →

See all 14 questions →