Learn AI by building it
PaperMan takes real AI research papers and turns them into coding challenges you solve in your browser. No setup. No fluff. Just you, the paper, and a code editor.
The Problem
Reading a paper doesn't mean you understand it
You can read "Attention Is All You Need" five times and still not be able to write a multi-head attention layer from scratch. Tutorials skip the hard parts. Courses move too slowly. And copying someone's GitHub repo teaches you nothing.
Our Solution
Build the paper, function by function
We break down each paper into small, testable functions. You write the code, our tests check if it's correct. When you're stuck, we show you the solution. By the end, you've built the entire architecture yourself.
How it works
1. Pick a paper
Choose from our library of AI research papers. Each one has been broken down into step-by-step coding tasks.
2. Write the code
We give you the function signature and a description. You fill in the implementation using our built-in code editor. No local setup needed.
3. Pass the tests
Hit "Submit" and we run your code against hidden test cases. Pass all the tests and you've proven you understand the concept — not just on paper, but in practice.
Why PaperMan?
Always up to date
Our AI agents automatically process new papers from ArXiv, so you get fresh challenges from the latest research.
No setup required
Write and run code right in your browser. We handle the Python environment, dependencies, and test execution.
Learn by doing
You don't memorize formulas — you implement them. Every function you write builds real understanding of how these models work under the hood.
Built for engineers
Whether you're a student, a researcher, or an ML engineer at a company, this is the fastest way to go from "I read the paper" to "I can build it."