Inspiration

For 1 million people to make good reproductive and fertility health at work available for 1 million men and women in the UK

What it does

It allows employees to get answers to questions they have regarding their reproductive and fertility health in a confidential way

How we built it

Python back-end using Google AI Agent ADK , Streamlit front-end, Pinecone Vector store, Supabase Managed SQL DB

Challenges we ran into

Achieving memory, persistent and personalised storage of chats. Inconsistent formatting of output and citations by the agent eg it cited the internal documents differently from those obtained externally

Accomplishments that we're proud of

Making it work! Integrating 3 different models - 2 Gemini models and 1 Open AI model into our solution using Open Router and Lite LLM

What we learned

Lots about the Google ADK which we're looking to leverage on other projects, multi-package implementation in Python, the uv package manager, Managed Cloud DBs

What's next for Policy Pulse

We will build out the solution to be more robust, more accurate and to give a better user experience. We will have extensive evals

Built With

Share this project:

Updates