Healthcare Readmission Risk MLOps Pipeline
MLOps-focused healthcare pipeline that turns a notebook model into a reproducible training, evaluation, and inference system.
Converted a Jupyter notebook diabetes readmission model into a production-style, reproducible MLOps pipeline with experiment tracking, config-driven runs, and a Dockerised FastAPI /predict service.
Key Outcomes
Context
Problem and Context
Notebook-based healthcare models are difficult to reproduce, operationalize, and evaluate reliably, especially when severe class imbalance makes minority-risk detection the real business objective.
Approach
Approach and Architecture
A config-driven 9-stage MLOps pipeline that transformed a diabetes readmission notebook into a reproducible training, evaluation, and inference system with MLflow, W&B, Hydra, GitHub Actions, Docker, and a FastAPI /predict service.
Implementation
Implementation Details
Ingestion -> preprocessing -> feature engineering -> modelling -> evaluation -> inference across a 9-stage pipeline. Hydra orchestrates repeatable runs, MLflow and W&B track experiments and artifacts, GitHub Actions automates checks, and Dockerized FastAPI exposes synchronous inference patterns for real-time and batch-ready use.
Results
Results and Tradeoffs
This project is presented as a concise technical overview rather than a full-length narrative case study.
Explore More
Related Projects
Browse adjacent work from the same archive group or jump back to the project archive.
Applied ML
Market Intelligence System
Financial intelligence pipeline blending transformer sentiment, quantitative signals, and multi-agent synthesis into structured risk reports.
Open case studyMLOps
Real-Time Fraud Detection Pipeline
Streaming fraud detection pipeline with online inference, drift checks, and dashboarded operational monitoring.
Open case studyAgentic Systems
LangGraph Autonomous Task Agent (Sidekick)
Stateful task agent that combines browsing, tool use, memory, and evaluation loops for multi-step execution.
Open case study