MLOps
MLOps Project: Road Accident Prediction
Containerized MLOps system for predicting road accidents – from data ingestion and feature engineering to production-grade REST API serving and an interactive Streamlit frontend.
In the road accident prediction project, I implemented a complete MLOps setup that covers the full lifecycle from data ingestion and feature engineering to model training and serving. The architecture uses containerized microservices (Auth, Data, Train, Predict) behind an Nginx API gateway, with MLflow for experiment tracking and model registry and DVC for data versioning. A Streamlit frontend allows users to easily input scenarios and predict accident severity, while two external APIs for geolocation and weather data populate feature fields quickly and accurately. With Docker Compose and k3s manifests, CI/CD pipelines, and monitoring via Prometheus and Grafana, the project demonstrates how to run ML models reproducibly and operate them reliably at production scale.
