Founding Engineer Take-Home

The Context Engine

Build a system that analyzes a GitHub repository and generates structured context for AI coding agents. Benchmark it against DeepWiki. Deploy it. Beat it.

Start the assignment

What you'll build

A context engine — a tool that accepts a GitHub repo URL and produces the structured information an AI coding agent needs to work effectively in an unfamiliar codebase. Project summary, architecture map, conventions, key files.

You'll build an eval framework that benchmarks your output against DeepWiki, then deploy it to a live URL and record a video walkthrough.

How it works

01

Start the clock

Fill out the start form. You'll get a template repo with the full assignment, API references, and suggested eval repos. Your 6-hour window begins when you submit.

02

Build & benchmark

Build a context engine that analyzes GitHub repos. Create an eval framework and benchmark your output against DeepWiki. Iterate until you're winning.

03

Deploy it

Ship your context engine to a live URL. Someone should be able to visit your site, enter a repo URL, and get results.

04

Show your work

Write a ~1 page walkthrough of your decisions. Record a 5-10 minute video demoing the product, codebase, and eval results.

Rules

6 hours maximum

Start form to submission form. If you're not done, submit what you have. We'd rather see how you prioritize under a constraint.

AI tools encouraged

Claude Code, Cursor, Copilot, agents — use whatever your normal workflow is. We care about the result and your methodology.

Your choice of stack

Language, framework, and approach are up to you. Use what makes you productive.

Ship early, iterate

Get something working and deployed before polishing. The progression from “working” to “better than DeepWiki” is what we want to see.

What we're evaluating

AI workflowHow you leverage AI tools to build — methodology, not just usage
ArchitectureRight-sized abstractions, good tradeoffs for the time
Eval thinkingCan you define "good", measure it, and reason about results?
Product instinctIs the output useful or just technically impressive?
ShippingIs it deployed? Does it work? Can we use it now?
CommunicationThe writeup and video reveal understanding code can't

Ready?

The full assignment, API references, and suggested eval repos are in the template repo.

Start the assignment