PolicyInsight

Inspiration

Policy documents are long, dense, and hard to navigate. I wanted to turn static policy text into structured, queryable data instead of something people skim and forget.

What It Does

PolicyInsight ingests policy documents and enables clause-level semantic search, structured requirement extraction, and risk tagging through a backend API.

How I Built It

Built with Java + Spring Boot, deployed on Cloud Run, powered by Vertex AI for embeddings and structured extraction. Retrieval blends semantic similarity with lexical scoring:

Challenges

Reducing hallucinations, handling messy PDFs, and maintaining low latency under load.

What I Learned

Precision > generic summaries. Strong guardrails and observability are critical for reliable AI systems.

Built With

• git
• github
• java
• powershell
• python