The situation
The fintech (around 350 people, around USD 4 billion in annual transaction volume) is regulated by the Monetary Authority of Singapore (MAS) and must show explainability and fairness diligence on any model touching credit before production roll-out.
Your task
Produce a SHAP-based explainability + fairness toolkit for a credit-risk model that satisfies the compliance team's model-risk requirements.
The brief
You start with
You receive a trained XGBoost credit-risk model (binary default prediction), the training feature schema (38 features), and a held-out 10,000-sample test set with labels.
What you do
- Compute global SHAP feature-importance plots, beeswarm + dependence plots, and per-decision local explanations for the test set.
- Run a fairness slice on default-rate predictions across 6 SME industry categories (services, retail, manufacturing, logistics, hospitality, agri).
- Produce a one-page explanation card template a credit officer can paste into the review file.
- Write a 2-page compliance memo with caveats on SHAP's limitations.
What you will practice
- Apply SHAP for global and local model explanation on tabular data
- Run a fairness slice analysis across protected-ish subpopulations
- Translate explanation outputs into operator-friendly artifacts
- Communicate XAI limitations to a compliance audience
A suggested way in
- Use TreeSHAP via the shap library for fast exact SHAP values on XGBoost
- Be explicit about base value vs. raw probability in local explanations; this trips up reviewers
- Report both predicted default rate AND calibration per slice; SHAP doesn't catch miscalibration
Skills you will practice
Shap, Interpretability, Fairness analysis, Model validation, Xgboost, Python
Where it leads
Data Scientist, AI Safety Researcher, Machine Learning Engineer
How it is assessed
- Explanation quality30%
Global and local SHAP outputs are correct and clearly visualized.
- Fairness slice analysis25%
Per-industry default-rate and calibration are reported with confidence intervals.
- Operator-card design20%
Explanation card is usable by a credit officer in under 60 seconds.
- Compliance memo25%
Memo addresses SHAP's known limitations (correlated features, off-manifold inputs) honestly.
Resources
- PythonTool
Standard language for SHAP and downstream analysis.
- shapTool
Reference library for SHAP values with TreeSHAP for XGBoost.
- PandasTool
Slice the test set by industry and compute fairness metrics.
- MatplotlibTool
Render dependence plots and per-slice calibration curves.
- FigmaTool
Design the one-page explanation card template.
- NotionTool
Author the 2-page compliance memo.
Finish it. Earn this.
A verified certificate with your name. Anyone can check it.
Time to practice.
Solve this challengeMore like this.
- Senior AccountantIFRS 16 Lease Accounting for a Renewable Energy Firm
- Software EngineerODE Integrator Benchmark for a Pharma PK/PD Pipeline
- Data ScientistInterpretable-by-Design GAM for an Insurer's Claims Triage
- Data Protection OfficerData Protection Impact Assessment for a Fintech App
-
IFRS 16 Lease Accounting for a Renewable Energy Firm Senior Accountant You are a senior accountant at SunPower Renewables. The company is adopting IFRS 16 for the first time. You are given a sample lease for a wind turbin ifrs 16lease accountingright of use asset Open challenge → -
ODE Integrator Benchmark for a Pharma PK/PD Pipeline Software Engineer Use SciPy's solve_ivp + SUNDIALS (via scikit-sundae or diffeqpy) to benchmark: RK45 (baseline), LSODA, BDF (CVODE), Radau, and Rosenbrock. Run on thre numerical odesstiff integratorsscientific python Open challenge → -
Interpretable-by-Design GAM for an Insurer's Claims Triage Data Scientist You receive an anonymized claims dataset (around 60,000 claims, target: log reserve), a feature schema (22 features), and an existing LightGBM baselin generalized additive modelsebminterpretability Open challenge → -
Data Protection Impact Assessment for a Fintech App Data Protection Officer You are the data protection officer (DPO) for the fintech. Perform a DPIA for the new AI-driven loan feature, following ICO guidelines. Identify at le gdpr compliancedata protection impact assessmentrisk management Open challenge →