Repo: github.com/louayamor/retinaxai ·
61 commits·41 open issues·TypeScript 74% · Python 17%
RetinaXAI is a production-grade monorepo combining MLOps, LLMOps, and a clinical web interface for automated diabetic retinopathy detection and medical report generation from OCT retinal scans.
Services:
| Service | Responsibility |
|---|---|
mlops-service |
Preprocessing, training (EfficientNet/XGBoost), MLflow tracking, DVC versioning, DagsHub integration |
llmops-service |
RAG pipeline (LangChain + ChromaDB/FAISS), local LLM inference via Ollama, PDF report generation |
backend-service |
FastAPI REST API, JWT auth, patient and admin management, service orchestration |
frontend-service |
Next.js + TypeScript + Tailwind CSS, clinician and patient dashboards |
infra |
Docker Compose, Kubernetes manifests, NGINX, GitHub Actions CI/CD, Prometheus + Grafana |
Stack:
ML / Training
Infra & DevOps
Frontend & Mobile
Databases
| Project | Description | Stack |
|---|---|---|
| RetinaXAI | Multimodal AI platform for diabetic retinopathy — MLOps + LLMOps + clinical UI | EfficientNet · LangChain · RAG · FastAPI · Next.js · Docker · K8s |
| Reasona | Lightweight GPT-style reasoning model with full MLOps training cycle | PyTorch · MLflow · LoRA · QLoRA · FastAPI |
