2025Project Leader · Backend & ML Developer
Wound-Care Recommendation Chatbot
A chatbot that classifies a wound from a photo with a CNN, then recommends first-aid steps via a rule-based system.
WoundCare
01 — Problem
What I set out to solve
- People often don't know how serious a wound is or how to treat it. A photo alone isn't enough — good advice also depends on context like age and other factors, so a pure image classifier would give confident but incomplete recommendations.
02 — Approach
How I built it
- Trained a CNN to classify a wound from an uploaded photo (e.g. abrasions) with a confidence score.
- Layered a rule-based system on top, folding in parameters like wound class, age, and sex to shape the advice.
- Wrapped the whole thing in a chatbot so the interaction feels like asking, not filling a form.
- Led the project and owned the backend and ML side of the build.
03 — Outcome
What it achieved
- An end-to-end pipeline from photo → CNN prediction → contextual, rule-based first-aid recommendation.
- A conversational interface that returns a prediction, a confidence level, and next-step guidance.
- Combined deep learning with explainable rules so the recommendation isn't a black box.
Stack
Built with
- Python
- TensorFlow
- HTML
- CSS
- JavaScript