Employee Attrition Risk Prediction Microservice

This end-to-end machine learning service predicts employee attrition risk using Gradient Boosting Survival Analysis (GBSA). Designed for HR teams, it identifies at-risk employees with interpretable explanations, enabling targeted retention efforts.

View on GitHub Try it Yourself

Technical Implementation

1. Machine Learning Core

Survival Modeling:

Trained Gradient Boosting Survival model to estimate time-to-attrition probabilities
Evaluated with concordance index (C-index) achieving [X]% on test data
SHAP (Kernel Explainer) for feature importance visualization
Interactive survival curves for top-N high-risk employees

2. Deployment & Infrastructure

Backend

FastAPI microservice
Docker
Google Cloud Run

Frontend

Streamlit dashboard

Data Pipeline

Google Cloud Storage (GCS)
CSV upload processing

3. Future Roadmap

Synthetic data generation for scenario testing
Auto-retraining pipeline with MLFlow and Databricks

Skills Demonstrated

Advanced ML

Survival Analysis
C-index Evaluation
SHAP Explainability

MLOps

FastAPI
Docker
Cloud Run

Data Engineering

GCS Integration
Auto-retraining

Impact

Reduces turnover costs through early risk identification
Explainable service builds HR trust in predictions
Production-ready system demonstrates full ML lifecycle ownership