Back
Eye Cancer Detection Model
Year
2025
Tech & Technique
Python, TensorFlow, EfficientNetV2, React, Node.js, FastAPI, Keras, REST APIs
Description
A production-grade, three-tier AI platform for automated ocular cancer detection using state-of-the-art deep learning.
Key Features:
Technical Highlights:
Key Features:
- 🧠 Trained EfficientNetV2-B3 achieving 97% accuracy and AUC 0.9971
- ⚡ Real-time inference pipeline with 150ms model latency and 350ms total system latency
- 🔒 Secure patient tracking with unique, image-linked identifiers
- 🗂️ Immutable prediction history tied to patient records
- 🌐 Full three-tier architecture: React frontend, Node.js auth service, FastAPI inference service
Technical Highlights:
- Containerized microservices enabling scalable inference
- Optimized preprocessing with OpenCV and NumPy for efficient GPU workloads
- REST-based communication between Node.js and FastAPI layers
- Integrated caching strategies to reduce redundant inferences
My Role
Full-Stack ML Engineer
I led the complete development and deployment of the platform:
I led the complete development and deployment of the platform:
- 🧪 ML: Trained & fine-tuned EfficientNetV2-B3 with TensorFlow/Keras
- ⚙️ Backend: Built FastAPI inference microservice with GPU-enabled acceleration
- 🛡️ Node.js Layer: Implemented patient tracking, access control, and prediction logging
- 🎨 Frontend: Developed the medical dashboard using React
- 🚀 Deployment: Containerized and deployed microservices with scalable API endpoints