AI · 2025
Skin Condition Detector – AI Backend
Express.js API that runs a YOLOv8 model trained on DermNet through ONNX Runtime, deployed serverless on Vercel.
PRJ-0002 · 202501 · Problem
Serving a computer-vision model usually means keeping a GPU server running. The goal was an inference API for a YOLOv8 skin-condition model that runs on serverless infrastructure and treats every upload as untrusted input.
02 · Approach
An Express.js service that loads a YOLOv8 model trained on the DermNet dataset and exported to ONNX, and runs it with ONNX Runtime inside a Vercel serverless function. Sharp normalises and resizes uploads before inference, and Helmet.js hardens the HTTP surface.
03 · Outcome
Model inference served from a serverless function, with no dedicated GPU infrastructure to run or pay for.
