Dermobot

Frontpage :)

💡 Inspiration

The average wait time to see a doctor in dermatology takes 30 weeks. With our app, you can get accurate results within seconds, making preventative healthcare more accessible than ever using machine learning and AI.

🧠 What it does

Dermobot lets users upload a photo of a skin lesion or concern and predicts the most likely skin condition based on a deep learning model. It also provides personalized, conversational advice and information from a medically-informed AI chatbot — all in one web app.

🛠️ How we built it

Trained a PyTorch ML model on 20,000 real skin disease images such as skin cancer and eczema
Built with React and Next.js for the frontend, and FastAPI for the backend to serve ML predictions and chatbot responses.
Designed a full-stack web app (React + Next.js) for image upload and user interaction
Used MongoDB to store metadata such as image submissions and user session data
Integrated an AI chatbot (Gemini API) using Retrieval Augmented Generation. We fine-tuned prompts and grounding data using trusted sources like MEDAI, WebMD, NIH, and Mayo Clinic.

🚧 Challenges we ran into

Connecting the ML model to a live web app via FAST APIs with minimal latency
Handling image upload and preprocessing in the backend
Designing a chatbot that’s helpful, accurate, and ethically responsible

✅ Accomplishments that we're proud of

Built a working prototype that integrates ML inference and chatbot communication
Designed a user-friendly UI with a real medical use case
Used RAG (Retrieval-Augmented Generation) framework to create a more knowledgeable chatbot

📚 What we learned

How to deploy ML models in production-ready APIs
Integrating AI into a healthcare context while prioritizing user safety and clarity

🌐 What's next for Dermobot???

Training the ML model on more skin diseases and body symptom datasets
Optimize run time and scale predictions by leveraging better GPU access for real-time performance under heavier loads
Improve chatbot accuracy and add multilingual support for broader accessibility