PersonalizedMedEdu

📌 Inspiration

Healthcare communication is often too complex for patients to fully understand. Many people leave medical appointments feeling confused about their condition and treatment plan, leading to poor adherence, anxiety, and misinformation. We wanted to create a personalized AI-driven solution that simplifies medical education, making it accessible, engaging, and tailored to each patient's needs.

💡 What it does

Our AI-powered system generates personalized medical education materials in text and video format. It adapts to the patient’s:
✅ Condition/Disease (e.g., Diabetes Type 2, Hypertension)
✅ Treatment Plan (e.g., Metformin, Lifestyle changes)
✅ Literacy Level (Simple, Intermediate, Advanced)
✅ Language (English, Spanish, etc.)
✅ Preferred Format (Text, Video, Audio)
✅ Emotional Sensitivity (Anxiety, Fear, Neutral)
Outputs: • Text summary (tailored to literacy level)
• AI-generated video (animated with voice-over)
• Q&A section (auto-generated from trusted medical sources)

🏗 How we built it

We combined LLMs, text-to-speech, and video generation APIs into a seamless AI-powered pipeline:
1️⃣ Frontend – Built using React.js for a simple and intuitive UI where users enter their medical details and preferences.
2️⃣ Backend – Developed with FastAPI, integrating multiple AI models and APIs.
3️⃣ LLM (Hugging Face - Mistral 7B) – Generates personalized text summaries based on patient inputs.
4️⃣ Silero TTS API – Converts generated text into natural-sounding speech.
5️⃣ D-ID API – Creates AI-generated videos with voice-over and animations.
7️⃣ Database – Uses PostgreSQL to store patient preferences and generated content.

🚧 Challenges we ran into

✅ Personalization complexity – Creating content that adapts to different literacy levels, languages, and emotional states required fine-tuning.
✅ Seamless API integration – Combining multiple AI-powered services (LLM, TTS, and video generation) while maintaining low latency.
✅ User experience design – Making a simple yet effective UI that works for diverse audiences.

🏆 Accomplishments that we're proud of

🚀 Successfully built a MVP for an AI-powered system that personalizes patient education across text, audio, and video.
🎯 Integrated multiple AI models (LLM, TTS, Video) into a cohesive and functional pipeline.
🌍 Enabled multilingual medical education to make healthcare knowledge more accessible.
📊 Created an adaptive Q&A system that automatically pulls relevant medical FAQs.

📖 What we learned

🔹 How to optimize AI-generated medical content for clarity and accessibility.
🔹 The importance of verifiable medical sources in AI-driven healthcare solutions.
🔹 Challenges of API orchestration when dealing with multiple AI services.
🔹 How to fine-tune LLMs for structured medical responses.
🔹 Designing an inclusive and user-friendly interface for a diverse audience.

🚀 What’s next for Untitled

🔜 Enhancing video generation – Improve animations and add more interactive elements.
🔜 Expanding language support – Covering more languages and dialects.
🔜 Better personalization – Fine-tune AI to better adapt to emotional sensitivity and literacy levels.
🔜 Mobile app development – Launching a dedicated mobile app for accessibility.
🔜 Integration with EHR systems – Allow doctors to generate personalized education materials directly from electronic health records.

Built With

• d-id
• fastapi
• huggingface
• mistral7b
• postgresql
• python
• react
• typescript