Charaka AI - Medical Assistant

Features

• Medical LLM: Fine-tuned Huatuo-GPT model specifically for medical domain expertise
• Multimodal Processing: Analyzes patient video and audio for symptom detection
• Efficient Fine-tuning: Implemented QLoRA (Quantized Low-Rank Adaptation) for optimized model training
• Diagnostic Assistance: Offers evidence-based suggestions and medical information
• Multilingual Support: Handles consultations in multiple languages

Technical Architecture

Model

Charaka AI uses a fine-tuned version of Huatuo-GPT, a specialized medical language model. The model was further optimized using QLoRA to:

• Reduce memory requirements during fine-tuning
• Maintain high performance while improving efficiency
• Enable deployment on consumer-grade hardware

Multimodal Pipeline

Video/Audio Input → Preprocessing → Feature Extraction → Medical LLM → Response Generation

Contributing

Contributions are welcome! Please check our Contributing Guidelines.

License

This project is licensed under the MIT License.

Acknowledgments

• Huatuo-GPT team for the base medical model
• QLoRA research team for quantization techniques
• Our medical advisors for domain expertise