About
What DeepForge is, who built it, and why.
What is DeepForge?
DeepForge is a practice ground for the math behind machine learning. Every problem asks you to implement a fundamental primitive from scratch — no sklearn, no torch, no imports beyond the standard library. Your code runs in a real Python interpreter (Pyodide) inside your browser, executed against real test cases.
The goal is fluency. If you can write a softmax by hand, you understand softmax. If you can derive the gradient of an MLP, you understand backprop. DeepForge is the place to do that work.
How it works
- 01Pick a problem from any category.
- 02Read the spec, then fill in the function body in the editor.
- 03Press Run. Pyodide loads once (~5s), then runs your code against every test case.
- 04Pass all tests to mark the problem solved. Your code and progress save to localStorage — no account needed.
Writing
Engineering notes on how DeepForge is built — the in-browser Python runtime, the problem pipeline, and the math under the hood.