MalamaAI – Skin Disease Detection
MalamaAI is a machine learning-powered application designed to recognize a variety of skin diseases using state-of-the-art computer vision techniques. The name “Malama”, meaning “to care for” in Hawaiian, reflects the project’s goal — to extend the reach of early detection and care through accessible technology.
At the heart of MalamaAI is a custom AI pipeline that builds upon a fine-tuned version of DinoV2, a robust self-supervised vision transformer. To further enhance its diagnostic capabilities, we integrated the Llama 3.370b language model, enabling the system to generate human-readable, context-aware reports based on image analysis results.
Features:
- Interactive Frontend: Built using Next.js for a responsive and smooth user experience.
- Scalable Backend: Flask-powered API server that efficiently handles image uploads and model inference.
- Enhanced Model Stack: Fine-tuned DinoV2 for image classification, combined with Llama 3.370b for natural language explanations and guidance.
MalamaAI aims to support users with accessible, AI-powered insights into their skin health. While not a substitute for medical professionals, the tool is a step toward democratizing health diagnostics using machine learning.
Technologies Used:
- DinoV2 (fine-tuned)
- Llama 3.370b
- Next.js
- Flask
- RESTful API integration