What it does CodaLab Competitions is a self-hostable web platform for organizing machine-learning competitions and sharing data-science worksheets. Researchers can participate in existing challenges or spin up private instances via Docker Compose. The project has been around since 2013, with stewardship passing from the Outercurve Foundation to Université Paris-Saclay.

The interesting bit The README’s most prominent feature is its own obsolescence notice: Codabench, billed as the “next-gen” replacement, is already live and actively promoted. This makes the repository something of a historical artifact with maintenance mode vibes—still functional, but clearly not where the maintainers’ energy is going.

Key highlights

• Docker-based deployment; docker-compose up -d gets a local instance running
• Hosted instance available at codalab.lisn.fr for those who don’t want to self-host
• Academic pedigree: published in JMLR (2023), with formal citation format provided
• Apache 2.0 licensed, with copyright chain from 2013–2021
• Companion “worksheets” feature for collaborative data-science notebooks

Caveats

• Linux-only quick-install instructions; macOS/Windows users are left to figure it out
• Documentation is scattered across a GitHub wiki and Google Groups forum
• The “next-gen” successor is already out, which raises questions about long-term support for this version

Verdict Worth a look if you’re running an academic lab that needs a fully open-source, self-controlled competition platform with formal citation cred. Skip it if you want modern UX, active feature development, or don’t enjoy maintaining decade-old Python web apps.