← all systems · case file 03 · score_churn() · 10k-Record XGBoost · Voice Briefing
ARIA
Predicts which customers will leave, then explains why in plain words and voice.
Summary
Bank Churn Intelligence: a decision-support prototype for banking teams. An XGBoost churn classifier, trained on a 10,000-customer dataset and served through FastAPI, returns a churn probability, a three-tier risk class and the top feature importances. Llama 3.3 70B on Groq turns each prediction into plain-English retention advice, and ElevenLabs reads it aloud.1GitHub repository
| Value | What | Method | Source |
|---|---|---|---|
| 10k | Customer records trained on | Rows in the bank-customer training dataset, per the README. | [1] repo |
| 3 | Actionable risk tiers | Churn-probability bands from the README: Low ≤35%, Medium 35–65%, High >65%. | [1] repo |
| 5 | Top predictive features surfaced | Global feature importances from the trained model, shown in the UI. They rank the model's drivers overall, not per customer. | [1] repo |
Problem
A raw churn probability isn't actionable on its own. Bank staff need a risk assessment they can understand at a glance and talk through with the customer.1GitHub repository
Architecture
XGBoost classifier behind FastAPI; three risk tiers; Llama 3.3 writes the summary, ElevenLabs reads it
01Imbalanced classification
XGBoost trained on 10,000 customer records, with SMOTE for class imbalance and one scaler shared by training and inference.
02Risk tiers and drivers
FastAPI serves real-time churn probabilities, grouped into Low, Medium and High risk, alongside global feature-importance rankings.
03Explanation and voice
Llama 3.3 70B writes plain-language explanations and next-step retention strategies; ElevenLabs speaks them.
Not claimed
- Is a bank running it in production?
- My sources call it a prototype, so I won't claim a deployment.
Sources
source [1] · Repository · ARIA
GitHub repository
github.com/vinayak533/bank-churn-intelligence-platform
