ActivityChecker

ML pipeline that predicts Strava activity patterns — whether a user will be active in the next 8 hours and how many kudos an activity will receive. Automated training, evaluation, and daily predictions via GitHub Actions.

2024Personal

Source Code

PythonMachine LearningDockerDVCGitHub Actionspersonal

// purpose

Wanted to build an end-to-end ML pipeline on real data — from Strava API ingestion to automated model selection and deployment.

// takeaway

Automating the full ML lifecycle is harder than training the model itself.

// stack

Pythonscikit-learnDockerDVCDagsHubGitHub Actions

// challenges

Automating model comparison so only better-performing versions replace the current one.

Building a reliable daily prediction pipeline without manual intervention.

Managing data versioning across training cycles with DVC.