Heal Veda

• 

• 

  system Architecture

Inspiration

Healthcare information is often confusing, especially when prescriptions are handwritten and when people mix allopathic medicines with Ayurvedic remedies without clear guidance. Heal Veda was built to simplify this gap using AI — focusing on understanding, not diagnosis.

What it does

Heal Veda is an AI-powered wellness safety platform that:

1. Scans printed and handwritten prescriptions using OCR
   
2. Explains medicines in simple, non-technical language
   
3. Provides basic medicine timing guidance (non-diagnostic)
   
4. Creates safety awareness for Ayurveda + allopathy combinations
   
5. Generates personalized herbal wellness plans
   
6. Stores interactions using a semantic memory system
   

The platform focuses on education and preventive care, not treatment.

How we built it

• OCR pipeline using EasyOCR
  
• Rule-based logic for medicine detection and timing
  
• Generative AI (Gemini Flash) for explanations and safety guidance
  
• Vector database (ChromaDB) for semantic memory
  
• Sentence Transformers for embeddings
  
• Flask backend with modular APIs
  
• Tailwind CSS for a clean, user-friendly interface
  

Challenges we ran into

• Inaccurate OCR from handwritten prescriptions
  
• Handling unclear medicine names safely
  
• Preventing AI from giving diagnosis or dosage
  
• Managing AI unavailability during high demand
  

These were handled using fallback logic and strict prompt constraints.

Accomplishments that we're proud of

• Built a real-world AI healthcare application
  
• Integrated OCR, LLMs, and vector databases end-to-end
  
• Designed a safe, explainable, non-diagnostic AI system
  
• Completed a full AI project during internship
  

What we learned

• Practical application of Generative AI in healthcare
  
• OCR + LLM pipeline design
  
• Semantic memory using vector databases
  
• Prompt engineering for safety-critical domains
  

What's next for Heal Veda

• Multilingual support
  
• Voice-based explanations
  
• Improved handwritten OCR
  
• Long-term user wellness memory
  

Built With

• apis
• cloud-services
• css
• databases
• flashllm
• flask
• frameworks
• geminiapikey
• html
• javascript
• llm
• platforms
• python
• restapi
• sentencetransformer
• tailwindcss
• vectordb
• webspeech