NLP Support Ticket Routing
Automated routing of support tickets using NLP classification.
Overview
Faster, more accurate routing improves customer satisfaction and efficiency.
Problem & Constraints
Manual ticket triage is slow and error-prone, delaying response times.
Data
50k+ tickets, multi-label, text, stratified split, handled class imbalance.
Approach
Baseline: TF-IDF + Logistic Regression → fine-tuned BERT → error analysis.
Evaluation
Macro F1: 0.81, Routing accuracy: 92%
Results
Reduced triage time by 60%, improved first-response SLAs.
Production / Engineering
- Inference design: See repo for details
- Latency & cost: See highlights
- Monitoring: Monitoring ideas in repo
Demo
Demo coming soon.
Links
What I'd Do Next
See repo for future improvements and roadmap.