Overview
What this challenge is about.
You receive 18 months of transactions (around 12M rows) and seller-firmographic data. Define a defensible proxy label for default (e.g., a 60-day chargeback-or-dispute spike combined with payment-volume collapse), engineer features, train and compare 3 model families (logistic regression, gradient-boosted trees, and a small neural net), and evaluate with a temporal holdout. Deliverable is a tuned model, a model card with calibration plots, and a 4-page memo explaining what the model can and cannot do for the credit committee's first underwriting decisions.
The Brief
What you'll do, and what you'll demonstrate.
Ship a default-risk model with a defensible proxy label and an honest model card the credit committee can underwrite the first cohort against.
Earning criteria — what you'll demonstrate
- Engineer features from transaction time-series for a tabular ML model
- Compare model families honestly with a temporal (not random) split
- Calibrate probabilities and explain the difference between rank and probability
- Write a model card that supports a real-money decision
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.
Machine Learning Engineer
Building a tabular risk model end-to-end with proper temporal evaluation and a model card is the day-one job of a junior MLE at a fintech or risk-modeling team.
This challenge sharpens
- feature-engineering
- model-evaluation
- ml-pipelines
Data Scientist
Defining a proxy label, comparing model families, and writing the decision memo for a credit committee is the core data-science loop in lending.
This challenge sharpens
- model-selection
- calibration
- feature-engineering
Applied AI Scientist
Translating a modeling result into a real-money underwriting decision with an honest model card is exactly what applied AI scientists do in regulated industries.
This challenge sharpens
- model-evaluation
- calibration
- feature-engineering