The situation
The fintech (around 250 staff, around 5 models in production across credit and fraud) has been losing 3 to 5 engineer-weeks per quarter to bespoke deploys; a v0 platform that compresses that to days is the team's stated OKR.
Your task
Ship a v0 MLOps platform slice (tracking, registry, training pipeline, monitoring) for a fintech risk team's next model deployment.
The brief
What you do
- Across a 5-person team, ship (1) experiment tracking integrated into a sample model training job; (2) a model registry that promotes-by-tag; (3) a training pipeline orchestrated by a workflow tool (Prefect / Dagster / Argo); (4) a monitoring dashboard for one deployed model showing input drift and a target accuracy proxy; (5) a runbook describing the daily and weekly operations.
- Use a sample fraud-detection model on a public dataset as the test customer.
- Deploy to a small cloud sandbox.
- CI gates merges.
- Produce a 6-page architecture document plus a 30-minute risk-team handover.
What you will practice
- Architect a small but real MLOps platform across multiple components
- Operate an experiment-tracking + model-registry workflow end-to-end
- Implement input-drift and accuracy-proxy monitoring
- Hand a platform off to a domain team with a usable runbook
A suggested way in
- MLflow + Prefect is a low-friction combo for the v0; document the why
- Drift detection: start with a Population Stability Index on top features
- Runbook beats README — write what operators do at 9 AM Monday
Skills you will practice
Mlops design, Experiment tracking, Model registry, Monitoring, CI CD, Team collaboration
Where it leads
MLOps Engineer, Machine Learning Engineer, AI Solutions Architect
How it is assessed
- Platform completeness30%
All five components work and connect cleanly.
- Monitoring rigor25%
Drift and accuracy-proxy metrics are sound and dashboards are readable.
- Handover quality25%
Risk team can run the platform from the runbook unsupervised.
- Team engineering20%
Git history shows code review, CI, branching discipline.
Resources
- PythonTool
Glue across pipeline, registry, monitoring.
- MLflowTool
Battle-tested experiment tracking + model registry.
- PrefectTool
Orchestrate the training pipeline.
- StreamlitTool
Quick monitoring dashboard.
- GitHubTool
Pull requests, CI, and code review hygiene.
- NotionTool
Author the runbook and architecture document.
- LinearTool
Track the 6-week capstone backlog and ownership.
Finish it. Earn this.
A verified certificate with your name. Anyone can check it.
Time to practice.
Solve this challengeMore like this.
- Software EngineerBuild a Multi-Level Feedback Queue Scheduler in xv6
- Software EngineerReverse-Engineer a Stripped Binary to Restore Linking Symbols
- Software EngineerImplement a Distance-Vector Routing Algorithm
- AI EngineerEnd-to-End Lane Following on a Donkeycar Platform
-
Build a Multi-Level Feedback Queue Scheduler in xv6 Software Engineer Fork the xv6-riscv repository and implement MLFQ in proc.c with 3 priority queues, per-queue time slices (10ms / 20ms / 40ms), priority demotion on sl operating systemsschedulingc programming Open challenge → -
Reverse-Engineer a Stripped Binary to Restore Linking Symbols Software Engineer Receive the stripped shared object (Linux ELF, x86-64, ~280KB), a crash dump with 12 frames of indecipherable addresses, and 4 years of partial git hi reverse engineeringlinking and loadingmachine code Open challenge → -
Implement a Distance-Vector Routing Algorithm Software Engineer Implement distance-vector routing in Python: each simulated router maintains a distance vector, exchanges vectors with neighbors every 1 second of sim routingdistance vectornetwork simulation Open challenge → -
End-to-End Lane Following on a Donkeycar Platform AI Engineer Use the public Donkeycar Tub dataset (or collect about 30 minutes of driving on the simulator). Train a CNN-policy baseline (the Donkeycar default arc end to end learningimitation learningpytorch Open challenge →