%md
Inspiration
We have seen patients having difficulties understanding when to consult a physician or take home medication for their symptoms
What it does
Input: Patient Info : Age, Gender, Occupation, symptoms etc.,
Model Activities:
Model will get information around past diagnostics patterns Information around various treatment plans Suggest options for the treatment/next steps Model evaluation: For example : If the patient has backpain, It will suggest chiropractor sessions and recommendations on best way to sit and posture which can help subsiding the backpain
How we built it
Built Gen AI RAG app pipeline: GenAI model: databricks-llama-4-maverick llm RAG chain with question, context and chat history Log the RAG app to Mlflow
Dataset used Disease Prevalence Rates -> Demographic Physician Dictation Data : Family Medicine -> It has Office Visit transcripts from physicians across almost all US states. Source: https://marketplace.databricks.com/details/bc106eb9-e5dc-443c-b015-f7a9b6aa686f/Shaip_Physician-Dictation-Data-Family-Medicine All the datasets have been processed which included RecursiveCharacterTextSplitter and vector embeddings using GTEgte. Lastly them were saved in vector search index for online similarity search.
Challenges we ran into
- Healthcare data on Databricks marketplace were not working and struggled to find correct dataset
- Integrating with vector index table created
Accomplishments that we're proud of
- The code runs successfully
- Successfully implemented RAG chatbot model serving in databricks
What we learned
- Build RAG application with databricks genAI agent
What's next for Physician chatbot
- Integrate with over the counter prescription details for non life threatening symptoms
Built With
- databricks
- llama4
- marketplace
- python
Log in or sign up for Devpost to join the conversation.