codalab-competitions

Why we included this project

CodaLab Competitions is a self-hostable web platform for running machine learning competitions, the kind where organizers upload datasets, define scoring, and let participants submit predictions or full code that executes inside Docker containers. It has powered hundreds of real scientific competitions across fields like physics and computer vision, and it supports multi-phase workflows and leaderboards that track several scores at once. Research groups and institutions that want a Kaggle-style setup on their own hardware will find it useful, including attaching their own CPU or GPU workers for larger runs. The codebase is mature and well documented, and even though the maintainers now point to Codabench as the next generation, this project is still a good reference for how competition platforms are built.