-
-
Analysis tab
-
Home tab
-
Compare tab
Inspiration
Loan professionals spend countless hours manually reviewing complex credit agreements that can be hundreds of pages long. Extracting key terms like interest rates, covenants, and collateral requirements is time-consuming and error-prone. We built LoanLens to transform this tedious process into instant, accurate analysis using AI.
What it does
LoanLens is a desktop application that analyzes loan documents with AI. It automatically extracts key information like borrower details, loan amounts, interest rates, and covenants from credit agreements. It converts complex legal language into plain-English summaries that anyone can understand. The comparison feature highlights changes between document versions and assesses their impact. Users can test the app with pre-loaded sample documents or upload their own files for real-time analysis powered by Google Gemini.
How we built it
We built the frontend with React and Ant Design in a black and gold theme, packaged as a desktop app using Electron. The backend uses Node.js and Express with a JSON database for portability. Google Gemini AI handles document analysis and intelligent comparisons. We deployed the backend on Render as a web service and implemented automatic cleanup to prevent storage issues. Sample documents use pre-computed results for instant demonstration while user uploads trigger live AI processing.
Challenges we ran into
Finding PDF parsing libraries that work across platforms without native dependencies was difficult. We needed to balance instant demo results with real AI functionality, which we solved through smart caching. Integrating the Gemini API required testing different models and initialization patterns. Render's ephemeral storage meant uploaded files would accumulate, so we built automatic cleanup on server restart. We also had to prevent users from comparing sample documents with uploaded ones since samples don't have actual PDF files.
Accomplishments that we're proud of
We built a fully functional desktop application with professional UI and reliable performance. Our AI prompts extract structured data from unstructured legal documents accurately. The demo experience is seamless with instant sample loading while supporting real AI for uploads. The code architecture separates PDF processing, AI analysis, caching, and API routes cleanly. We successfully deployed both frontend and backend with proper environment configuration and automatic deployment.
What we learned
We learned how to structure AI prompts to get consistent JSON responses from complex documents. Building production-ready Electron apps taught us proper window management and asset handling. We understood cloud deployment tradeoffs between ephemeral storage and caching strategies. Balancing instant feedback with real AI capabilities in demos requires careful UX design. Navigating the Google Gemini SDK documentation improved our API integration skills.
What's next for LoanLens
We plan to add risk scoring based on covenant strictness and automated red-flag detection for unusual terms. Team collaboration features like annotations and shared document libraries would help organizations work together. Advanced AI capabilities could include custom extraction templates for different loan types and natural language queries. Integration with loan origination systems and automated covenant monitoring would embed LoanLens intelligence into existing workflows.
Log in or sign up for Devpost to join the conversation.