Elastic HealthCare – Verified Medical Responses

“Your trusted source for accurate, AI-verified medical information. Skip the hallucinations of language models and rely on verified answers, powered by fast search and Vertex AI capabilities.”

Comment

Inspiration

We noticed that while AI models are increasingly used for healthcare Q&A, they often hallucinate or provide inaccurate medical information. Patients and developers need a trusted, verified source of medical knowledge. Elastic HealthCare was inspired by the need for reliable, doctor-verified answers delivered quickly and efficiently using AI and advanced search technologies.

What it does

  • Ask medical questions and receive verified answers from doctors.
  • Search through a curated dataset of medical Q&A and documents.
  • Get AI-assisted insights powered by Vertex AI while ensuring factual accuracy.
  • Quickly find relevant information using Elasticsearch-powered fast search.
  • Present answers in a clear, structured, and interactive UI, including supporting references and grounding documents.

How we built it

  • Backend: Python, Elasticsearch for storing and searching medical documents, Vertex AI for AI-assisted grounding.
  • Frontend: Streamlit for interactive, user-friendly pages.
  • Data: Curated medical datasets, Q&A pairs, and verified references.
  • Integration: Vertex AI embeddings used to semantically match questions with answers, Elasticsearch reranking for relevance, and embedding search for grounding.
  • Deployment: Hosted with GCP Cloud Run, Elasticsearch, and cloud-based Vertex AI services

Challenges we ran into

  • Search relevance: Tuning Elasticsearch and Vertex AI embeddings to return the most accurate answers.
  • Performance: Keeping response times low while performing semantic searches and reranking.

Accomplishments that we're proud of

  • Built a fully functional AI-assisted healthcare search platform.
  • Successfully integrated doctor-verified data with Vertex AI embeddings and Elasticsearch search.
  • Implemented relevance reranking and grounding for higher accuracy.
  • Created a polished UI in Streamlit, with features like suggested questions, multilingual support, and easy-to-copy answers.
  • Ensured answers are trustworthy and traceable, reducing the risk of misinformation.

What we learned

  • The importance of grounding AI responses to verified sources to maintain trust.
  • How to combine search engines and AI embeddings effectively for healthcare applications.
  • Strategies for optimizing indexing and reranking of large document sets.
  • Best practices for Streamlit app design to create interactive, user-friendly dashboards.

What's next for Elastic HealthCare – Verified Medical Responses

  • Integrate more verified datasets from trusted medical sources.
  • Improve AI grounding and reranking to handle more complex queries.
  • Explore personalized insights for healthcare professionals and patients.

Built With

  • cloud-run
  • cloud-storage
  • elastic-vector-search
  • elasticsearch
  • gcp
  • gemini
  • gemini-embeddings
  • inference-api
  • kibana
  • streamlit
  • vertex-ai
Share this project:

Updates