Find terms that need attention
Review findings about payment, indemnity, termination, insurance, and other contract provisions, with supporting text and source references.
AI Application2026 Hatchathon · Explorers Track Winner
Contract Connection
A contract review assistant I built with Jenny Florey to help construction teams find important terms, review potential risks, and keep related documents together.
Built with Jenny Florey
Demo available on request
Making a complicated review easier to work through.
Construction contracts often come with exhibits, schedules, and other documents that need to be read together. Finding a payment term is one task. Understanding how it relates to the rest of the agreement takes more work.
Jenny and I built Contract Connection during the two-week 2026 Hatchathon. We took first place in the Explorers track. I led the technical architecture and development, working with Jenny on the problem we wanted to solve.
The goal was to give people a useful starting point for review: help them find the terms that need attention and return to the original documents to check them.
From a document package to a focused review.
Related documents are grouped into a case. The application extracts their text and uses AI to identify terms and potential issues for a reviewer to examine.
Compare related documents
Look for conflicting terms across the contract package and references to documents that may be missing.
Ask questions
Search the documents in plain language and review the passages returned with the answer.
Share the review
Export findings to PDF, Word, or Excel so the team can use them in its existing review process.
Different readers need different details.
I organized the output around the people involved in a contract review. An executive may need a summary of the main concerns. A project manager needs payment terms, deadlines, and a checklist. A legal reviewer needs the clause text and its context.
Keeping those views together lets people work from the same analysis without making everyone read the same report. Source references give reviewers a way to check findings against the documents.
AI can miss something or interpret it incorrectly. The findings support human review; decisions about the contract still belong with the people responsible for it.
What I built.
I built the application with Python, FastAPI, and PostgreSQL, connecting document processing, AI analysis, search, and report generation. It includes OCR for scanned documents and support for multiple AI providers.
If you would like to see how it works, contact me and we can arrange a demo.
If you are building something that needs domain judgment, careful systems thinking, or practical AI support, send me a note.