Overview
What this challenge is about.
You receive the 60-signal dataset and a short interview summary of risk analysts' beliefs about which signals influence which. Use a hill-climbing structure-learning algorithm with a Bayesian-Information-Criterion (BIC) score and 3 random restarts. Compare the learned DAG (Directed Acyclic Graph) to a constraint-based baseline (PC algorithm) and to the analyst-elicited DAG. Score on (a) chargeback-prediction held-out log-loss when used as a Bayesian network, and (b) Structural Hamming Distance to the analyst DAG. Deliver the comparison plus a 2-page note to the CRO calling out the top three edges that should change the risk team's mental model.
The Brief
What you'll do, and what you'll demonstrate.
Learn the structure of a probabilistic graphical model over fintech risk signals and identify edges that should change the team's mental model.
Earning criteria — what you'll demonstrate
- Apply score-based (hill-climbing + BIC) and constraint-based (PC) structure learning
- Compare learned DAGs quantitatively (held-out log-loss) and structurally (Hamming distance)
- Interpret graphical-model edges in causal-but-careful language
- Communicate probabilistic findings to a C-suite audience
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.
ML Researcher
Comparing score-based and constraint-based structure learning on real risk data is genuine applied-research work and a strong portfolio piece for ML research roles.
This challenge sharpens
- structure-learning
- bayesian-networks
- causal-modeling
Applied AI Scientist
Turning a learned graphical model into three concrete belief updates for the CRO mirrors the daily craft of applied AI scientists in regulated industries.
This challenge sharpens
- bayesian-networks
- executive-communication
- model-evaluation
Research Scientist
The structure-learning comparison plus bootstrap stability analysis is the level of rigor research-scientist roles in industry expect from a first project.
This challenge sharpens
- structure-learning
- causal-modeling
- model-evaluation