Rule-Based Intent Classifier for a Customer-Support Triage Bot
Overview
What this challenge is about.
Build a rule-based classifier in Python that runs ordered rules (regex + keyword + simple heuristics) against ticket subject + body. Use a hierarchical rule structure (high-precision rules first, fall back to broader matches). Add a confidence threshold: only auto-route above threshold; below it goes to human triage. Evaluate on 6 weeks of historical tickets (around 4,800, with human-labeled categories) — measure auto-route coverage, accuracy, misroute rate per category. Identify the 3 categories rules struggle on and recommend whether they need ML in v2. Deliver: Python classifier + tests, 6-page evaluation report, 3-page Zendesk integration spec, and recommendation memo.
The Brief
What you'll do, and what you'll demonstrate.
Build a rule-based intent classifier for support triage that auto-routes 75 percent of tickets with under 5 percent misroute rate.
Earning criteria — what you'll demonstrate
- Design a rule-based classifier with hierarchical precision/coverage trade-offs
- Set confidence thresholds for hybrid AI + human-in-the-loop systems
- Evaluate classifier performance with per-category metrics
- Recommend when to escalate from rules to ML defensibly
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.
Product Manager
PMs on support-tooling products need this rules-first lens to scope automation that ships in 2 weeks, not 2 quarters.
This challenge sharpens
- rule-based-systems
- algorithm-evaluation
- knowledge-representation