ML EngineeringClosing soon
Deploy it, then watch it drift
Serve a trained model, replay three weeks of traffic, and catch the drift we hid in week three.
- Time needed
- 1 week
- Replay
- 3 weeks traffic
The brief
Training is the easy half. Package the provided model behind an endpoint, instrument it, and replay three weeks of traffic — one of which contains a distribution shift. The challenge is whether your monitoring catches it before the scoring script tells you.
What you submit
- A containerised prediction endpoint
- A monitoring report over the replayed traffic
- The alert you would have sent, and when it would have fired
How to participate
- Register free: create your account and tell us a little about yourself.
- Pull the model artefact and the replay dataset.
- Serve, instrument, replay, and document what you caught.
- Submit; mentors check whether your alert timing holds up.
How it’s judged
- Drift detected and when
- Endpoint reliability
- Usefulness of the alert
What you walk away with
Feedback for every entry · Top 3 win up to 75% off a flagship seat
Resources
Everything you need, in one zipThe dataset, the full brief and the guides for this challenge.Register free to unlock