ContractIQ

AI-powered contract intelligence — upload PDFs and instantly surface renewal deadlines, risk scores, key clauses, and obligations before they bite you.

Comment

Inspiration

Small businesses and startups sign dozens of contracts — vendor agreements, SaaS subscriptions, MSAs — and then lose track of them in shared drives. Renewal dates get missed. Auto-renewals lock teams into bad deals. Risky indemnity clauses go unreviewed. We built ContractIQ because legal ops shouldn't require a legal team.

What It Does

ContractIQ is an AI-powered contract intelligence dashboard. You upload a PDF contract, and within seconds you get:

  • Extracted metadata — parties, start/end dates, total value, payment terms, renewal terms
  • Clause analysis — renewal, termination, liability, SLA, confidentiality, and indemnity clauses, each tagged by importance (LOW → CRITICAL)
  • Obligations — actionable to-dos extracted from the contract with due dates and responsible parties
  • Risk scoring — every contract gets a 0–100 risk score based on renewal proximity, missing metadata, financial exposure, and clause gaps
  • Alerts — automated reminders for approaching renewal windows, payment deadlines, and obligation due dates

How We Built It

  • Frontend & Backend: Next.js 15 (App Router) deployed on Vercel — all API routes live alongside the UI in a single project
  • Database: AWS Aurora PostgreSQL — stores all contracts, clauses, obligations, alerts, and organizations
  • ORM: Prisma — schema-first type-safe access to Aurora
  • AI Extraction: Google Gemini (via Gemini API) — given raw contract text, returns structured JSON with parties, dates, clauses, obligations, and risk factors. Falls back to a deterministic rule-based extractor if the API is unavailable
  • Auth: NextAuth.js — demo login for judges, plus Google and GitHub OAuth
  • File Storage: Vercel Blob for uploads in production

Challenges

  • Structured extraction reliability: Getting Gemini to return consistent, valid JSON across wildly different contract styles required careful prompt engineering and a robust fallback extractor
  • Aurora connectivity from Vercel: Configuring SSL (sslmode=require) and connection pooling (connection_limit, connect_timeout) to make Prisma work reliably with Aurora Serverless from Vercel's edge-adjacent functions
  • Risk scoring without ground truth: Building a risk model that feels meaningful without labelled training data — we ended up with a weighted factor system that scores renewal urgency, clause severity, and missing metadata independently

What We Learned

  • Aurora PostgreSQL integrates cleanly with Prisma — the postgresql:// connection string just works once SSL is correctly configured
  • Gemini's structured output mode dramatically improves extraction reliability over prompt-engineering alone
  • A great product demo matters as much as the code — building a seeded demo workspace with realistic contracts let us show the full value immediately

What's Next

  • Email/Slack notifications for expiring contracts
  • Multi-user workspaces with role-based access
  • Bulk CSV import for companies migrating existing contract data
  • Clause negotiation suggestions powered by Gemini

Built With

Share this project:

Updates