Inspiration
What it does
How we built it
Inspiration
Financial analysts and researchers often waste hours sifting through lengthy PDF reports, balance sheets, and filings just to extract a few key insights. I wanted to create something that makes this process faster, more accurate, and less stressful — especially for people who don’t have advanced technical skills.
What it does
MyFinDocGPT is an AI-powered, open-source web app that instantly analyzes, summarizes, and compares financial documents in multiple formats (PDF, DOCX, XLSX, HTML, TXT). Users can:
Upload one or two documents for side-by-side analysis.
Get concise summaries, key metrics, and structured tables.
Ask natural-language questions about the content.
Save hours of manual reading while improving accuracy.
How I built it
Frontend & UI: Streamlit for a fast, interactive interface.
AI Processing: Google’s Gemini API for advanced text understanding and summarization.
File Parsing: Python libraries like PyMuPDF, docx, and pandas for extracting and structuring content.
Deployment: Designed to run locally or in a containerized cloud environment.
Challenges I ran into
Handling large PDFs efficiently without hitting API token limits.
Ensuring accurate extraction from different document formats.
Creating a smooth UX that works even for non-technical users.
Accomplishments I’m proud of
Built a functional MVP that works with multiple file types.
Kept the project fully open-source so others can adapt it.
Delivered an interface that feels intuitive and fast.
What I learned
Best practices for integrating large language models into domain-specific tools.
Optimizing file parsing for different formats.
Designing AI tools with transparency and user trust in mind.
What’s next
Add multilingual document support.
Integrate charts and visual analytics.
Deploy a hosted version for public use.
Log in or sign up for Devpost to join the conversation.