FedDiagnose: Cross-Hospital Federated Diagnostic Modeling
A simulated multi-hospital federated learning platform for chest X-ray classification, keeping patient imaging data local to each institution.
Overview
FedDiagnose simulates five hospital sites training a shared chest X-ray classifier via FedAvg, each holding a distinct, non-IID slice of the ChestX-ray14 dataset. A FastAPI orchestration layer coordinates rounds, and each "hospital" runs in its own Docker container to emulate network isolation.
Architecture
A central Flower server aggregates client updates from five containerized clients; a lightweight monitoring dashboard tracks per-round validation AUROC per site and aggregate model performance, with optional differential-privacy noise injection at the client update step.
Tech Stack
Screenshots
Lessons Learned
Non-IID label distribution across simulated hospitals was the single biggest driver of slower convergence — far more than communication frequency. Adding differential privacy noise required re-tuning the learning rate schedule, since naive settings destabilized early rounds.
ls -la ./related-projects/
Edge-NIDS: On-Device Intrusion Detection for IoT
A quantized CNN intrusion-detection model deployed to an ARM Cortex-M microcontroller for real-time IoT traffic monitoring without cloud connectivity.
ReproBench: Automated Paper Reproduction Scorecard
A tooling suite that runs reproduced model checkpoints against held-out benchmarks and auto-generates a comparison scorecard against reported paper results.
ClinSumm: LoRA-Tuned Clinical Note Summarizer
A parameter-efficient fine-tuned language model that summarizes de-identified clinical notes into structured, clinician-readable briefs.