buildml started as a personal project by praash — frustrated that every ML tutorial either hand-holds through copy-paste code or throws you at research papers with zero scaffolding. There had to be a middle path.
The answer: implementation challenges. Read the paper, write the code, pass the tests. No gradients handed to you. No shortcuts. Real understanding.
Today buildml has over 2,400 builders working through 40+ challenges spanning everything from backprop to diffusion models. The mission hasn't changed.
Reading papers gives you the map. Writing the code forces you to take the journey. Bugs reveal misunderstandings that re-reading never would.
Libraries abstract away the important parts. When you implement attention in numpy, you understand every matrix multiply. That transfers to real work.
Good engineers are competitive. A public leaderboard makes progress feel real and creates a community of people who care about understanding.
The platform is free. Challenges are free. We're supported by sponsors who believe in better ML education, not paywalls.
Started buildml because I wanted to learn AI/ML by implementing research papers from scratch. I’m still figuring things out, but I’m excited to build this community with you.
Every challenge takes you from theory to code. The understanding that comes from debugging your own attention implementation is irreplaceable.
Pick a Challenge →