Applying Phase 3 skills
Instruction and application
In ProgressThis hackathon focuses on four key skills that you will practise by taking an ML solution from a working notebook to a monitored, communicated, and documented deployment:
- Skill 1: Deployment and scalable architecture
Designing and executing a deployment for an ML model into a production or test environment — considering accessibility, scalability, security, cost, version control, and rollback.
- Skill 2: Monitoring, drift detection, and adaptation
Building a monitoring plan for a live model, detecting model and data drift, and defining retraining triggers and proactive adaptation strategies.
- Skill 3: Lifecycle management and risk governance
Managing the full lifecycle of a deployed model — including retraining, versioning, security and compliance considerations, and criteria for decommissioning.
- Skill 4: Stakeholder communication and technical documentation
Communicating deployment decisions, risks, and value to diverse audiences, and producing comprehensive technical documentation to support handover and inclusive collaboration.
Applying skills to real-life scenarios
Building on the model you developed in Hackathon 2, this hackathon focuses on the final stage of the AI/ML lifecycle: getting the model into production, keeping it healthy, and making sure the people around it understand and trust it.
The scenarios reflect the kinds of decisions teams face when a model leaves the notebook — where architecture, monitoring, explainability, and communication become just as important as the underlying performance metrics.
Action item: Discussion
Directions: Share your thoughts in the chat or come off mute
- What does it take for a deployed model to be considered "safe to keep running" in production?
- If your model's accuracy starts drifting after two weeks in production, how would you know — and what would you do first?
- Why is stakeholder communication often the difference between a technically good model and one that actually gets used?