Overview
What this challenge is about.
Pick one priority feature group (recommend the 25 transaction-history features used by the fraud model). Define the offline source-of-truth (likely Snowflake or BigQuery), the online store shape (Redis or DynamoDB), the materialization job (Airflow + dbt or similar), and the read API contract. Implement a working slice end to end for those 25 features, including a parity test that proves offline == online for a sample of users. Deliver a 5-page design doc + the working slice + a migration plan for the other 17 models.
The Brief
What you'll do, and what you'll demonstrate.
Design and prototype a feature-store pattern that eliminates train/serve skew for the fraud model and shows a clear migration path for the other 17.
Earning criteria — what you'll demonstrate
- Distinguish offline vs online feature serving and the skew it causes
- Pick a feature-store shape (rolled-your-own vs Feast vs Tecton) with reasoning
- Implement a working materialization pipeline with parity tests
- Design a migration plan that respects existing model deadlines
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
MLOps Engineer
Designing and shipping a feature-store v1 is the platform-MLOps work that every fintech and consumer-AI team eventually hires for after their first train/serve skew incident.
This challenge sharpens
- feature-store
- airflow
- parity-testing
Data Engineer
Owning the offline-to-online materialization pipeline and dbt + Airflow plumbing is core data-engineering work on any ML-platform team.
This challenge sharpens
- data-pipelines
- airflow
- feature-engineering
Machine Learning Engineer
MLEs increasingly own feature definitions end to end; this challenge bridges modeling fluency into the platform side that ships features other models reuse.
This challenge sharpens
- feature-engineering
- feature-store
- data-pipelines