🚑 MedAssist AI – A Full-Stack AI Symptom Checker Using Cohere LLM & Machine Learning

🔍 Overview

MedAssist AI is a full-stack web application that leverages both Supervised Machine Learning and Generative AI (Cohere) to help users understand potential medical conditions based on their symptoms.

This project was developed as part of the Microsoft AICTE AI & Edunet Foundation Internship, and it's now deployed with a React frontend (Vercel) and FastAPI backend (Render).

🧠 What Does It Do?

• Users input symptoms, age, and gender.

• The system provides:

  • ML-based predictions using a Random Forest Classifier.

  • LLM-based care tips using Cohere AI.

• Toggle between ML and AI outputs with real-time results.

🎯 Problem Statement

With growing reliance on search engines for self-diagnosis, users often face fear or misinformation. MedAssist AI aims to:

• Simplify symptom analysis

• Offer reliable AI-driven suggestions

• Improve accessibility to basic healthcare insights

⚙️ System Development Approach

Data Collection & Preprocessing

• Public symptom-disease datasets

• Encoded inputs: age, gender, symptom vectors

• Text vectorization: TF-IDF

ML Model

• Type: Supervised Learning

• Algorithm: Random Forest

• Features: Age, Gender, Symptom Vectors

• Labels: Disease Categories

Generative AI

• LLM: Cohere API

• Used for generating contextual care tips

Deployment

• Frontend: Vercel (React + Material UI)

• Backend: Render (FastAPI)

🛠️ Tech Stack

📈 Evaluation

• Accuracy measured using cross-validation

• Cohere responses evaluated based on clarity, relevance

• Seamless toggle between ML/LLM improves user trust

🔮 Future Scope

• Add user authentication for medical history tracking

• Expand to multilingual support

• Integrate symptom severity scoring

👨‍💻 What I Learned

• Building and deploying full-stack AI applications

• Model training and vectorization techniques

• Using LLMs for real-world use cases

📌 GitHub Repo

👉 GitHub – MedAssist AI

🔗 Try it Live

👉 Live Site – Vercel