Back to projects
AI & Research Research paper
Secure ML Model Training & Prediction for Healthcare (FHE)
A fully homomorphic encryption (FHE) server architecture enabling ML inference on encrypted healthcare data using Concrete ML.
Overview
A research project exploring how machine learning models can train and predict on healthcare data without ever decrypting it. Built a server architecture around Concrete ML that performs inference directly on fully homomorphically encrypted inputs, preserving patient privacy end-to-end while still delivering usable predictions through a Streamlit interface.
Highlights
- Designed an FHE-based inference pipeline using Concrete ML, keeping patient data encrypted throughout.
- Built supporting data pipelines and visualizations with Pandas, Scikit-learn, and Matplotlib.
- Exposed the model via a Flask API with a Streamlit front-end for demonstration.
- Published with DOI: 10.13140/RG.2.2.15680.14083
Open to security research collaborations & freelance engineering work
Let's strengthen your security posture — or build something new.
Whether it's detection engineering, a compromise assessment, or a full-stack build — I'm always glad to talk shop.