ContractPulse

Inspiration

Every year, businesses lose thousands of dollars to auto-renewed contracts they forgot about. A missed cancellation window on a $50K SaaS license can blow an entire quarter's budget. We saw this happen repeatedly in B2B teams juggling dozens of vendor relationships — and realized there was no simple, intelligent tool to solve it.

ContractPulse was born from a simple question: What if your contracts could tell you when they need attention?

What it does

ContractPulse is a contract renewal intelligence platform that:

• Parses contracts with AI — Upload a PDF and Amazon Textract automatically extracts key terms (renewal dates, notice periods, payment terms, liability clauses)
• Scores risk automatically — Contracts are flagged as high/medium/low risk based on value, auto-renewal status, and time remaining
• Sends smart alerts — A daily cron job checks for contracts expiring in 30, 60, or 90 days and notifies the right people
• Visualizes spend — Dashboard with vendor spend analytics to identify cost-saving opportunities
• Supports multi-tenancy — Org-scoped data with role-based access for teams

How we built it

We chose a production-grade stack from day one:

• Frontend: Next.js 14 with App Router, React Server Components, Tailwind CSS, and shadcn/ui for a polished, accessible UI
• Database: Amazon Aurora PostgreSQL (Serverless v2) — scales to near-zero during inactivity while providing enterprise reliability
• ORM: Prisma for type-safe database access and easy migrations
• AI: Amazon Textract's AnalyzeDocument API to extract structured data from contract PDFs
• File Storage: Vercel Blob for contract document storage
• Auth: NextAuth.js with JWT sessions and middleware-based route protection
• Deployment: Vercel with automatic CI/CD from GitHub
• Alerts: Vercel Cron Jobs running daily to check renewal windows

The architecture follows a clear separation: Vercel handles compute and edge delivery, AWS handles data persistence and AI processing.

Challenges we faced

1. Prisma on Vercel — Vercel's dependency caching meant Prisma Client wasn't regenerated on deploy. Solved by adding prisma generate to the build script.
2. Special characters in database URLs — Aurora's generated passwords contain characters that break URL parsing. Had to URL-encode the connection string.
3. Multi-tenancy scoping — Ensuring every query is scoped to the user's organization required threading session context through server components and API routes.
4. Textract term extraction — Raw Textract output is unstructured blocks of text. We built a keyword-matching parser to identify relevant contract clauses with confidence scores.

What we learned

• Aurora Serverless v2 is genuinely impressive for startups — you get enterprise PostgreSQL that costs almost nothing at low traffic
• Next.js App Router with Server Components eliminates most of the client/server data-fetching complexity
• Prisma + Aurora is a powerful combination for rapid prototyping that doesn't sacrifice production readiness
• Building for multi-tenancy from day one saves massive refactoring later

What's next

• Email delivery via Amazon SES for real alert notifications
• Slack integration for team notifications
• Contract comparison (diff two versions)
• Bulk upload and batch Textract processing
• Team collaboration features (comments, approvals, workflows)

Built With

• amazon-aurora-postgresql
• amazon-textract
• next.js
• nextauth.js
• prisma
• react
• shadcn/ui
• tailwind-css
• typescript
• vercel
• vercel-blob