Hey Product Hunt! I'm Rocky, one of the people behind TrenTorch. I wanted to understand what actually happens inside loss.backward() instead of just importing it. So we built a tiny PyTorch-style framework from scratch in plain NumPy, then turned it into a course you can do in your browser with nothing to install. What you can do today: - Write the pieces yourself (tensors, autograd, CNNs, transformers) in a built-in code editor - Track your progress across the curriculum - Try the Problem of the Day if you want one small thing to build each day - Or clone the GitHub repo and run the same path from a CLI It's still growing. The curriculum on the site is being written, so some topics have more questions than others, and I'd rather tell you that up front. I'd love your help with one question: what's the one ML concept you've used but never really understood under the hood? Tell me and I'll prioritise it. Feedback, bugs and ideas are all welcome here, or at rocky@trentorch.com.

TrenTorch

Learn Frontier ML by building it yourself.

Learn Frontier ML by building it yourself.