Overview
What this challenge is about.
You receive 90 days of sensor logs (vibration, spindle temperature, coolant flow, ambient humidity), the maintenance log of 180 failure events labeled by root cause, and a short interview transcript with the lead maintenance engineer describing 8 known failure pathways. Structure a Bayesian network with nodes for sensors, latent tool-wear states, and 8 failure modes. Learn parameters (Conditional Probability Tables) from the data using maximum likelihood with Dirichlet smoothing. Validate via 5-fold cross-validation on held-out failure events; success is a top-1 root-cause accuracy above 65 percent versus a 35 percent baseline of the current threshold rules. Wrap inference in a simple Streamlit demo the engineer can open at the line.
The Brief
What you'll do, and what you'll demonstrate.
Build a Bayesian network that infers the most likely root cause of CNC machining defects from sensor data with materially higher accuracy than the current threshold rules.
Earning criteria — what you'll demonstrate
- Translate domain knowledge into a directed graphical model structure
- Learn Conditional Probability Tables (CPTs) from observational data with smoothing
- Perform exact inference (variable elimination) on a moderate-sized network
- Communicate probabilistic outputs in a way a non-statistician engineer can act on
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
Designing, training, and shipping a Bayesian network behind a usable internal tool is exactly the kind of probabilistic-modeling work MLEs do at industrial customers where deep learning is overkill.
This challenge sharpens
- bayesian-networks
- python
- model-evaluation
Data Scientist
Working a noisy manufacturing dataset into a calibrated, interpretable probabilistic model is high-leverage data-science work for any industrial employer.
This challenge sharpens
- probabilistic-inference
- parameter-learning
- model-evaluation
Applied AI Scientist
Translating domain-expert interviews into a structured graphical model and validating it against real failures mirrors the daily craft of applied AI scientists in industrial AI.
This challenge sharpens
- bayesian-networks
- structured-modeling
- probabilistic-inference