Insightly AI

Insightly AI turns complex annual reports into instant business insights. Just upload a PDF and ask anything—our AI answers it in seconds. Built with passion by Rachana & Manas.

Comment

🎯 Project Title: Insightly AI — Transforming Annual Reports with AI

👩🏻‍💻 Rachana Srikruthi Venkumahanti 👨🏻‍💻 Manas Reddy

📘** About the Project:**

Insightly AI is an intelligent web platform designed to simplify the analysis of dense and lengthy annual reports using Artificial Intelligence. It allows users to upload any annual report in PDF format and then ask natural language questions like “What are the major financial risks?” or “What does the report say about Series A funding?”

Our platform parses the document, retrieves relevant context, and generates smart, detailed responses using Cohere’s large language model.

✨ Core Features:

🔹 Drag-and-drop annual report (PDF)

🔹 AI-generated Q&A on financial performance, strategy, risks, and more

🔹 Suggested question prompts based on content

🔹 Upload history with persistent access

🔹 Mobile & desktop responsive UI

🔹 Integrated error handling and CORS management

📊 Tech Stack:

Frontend: React (Bolt low-code platform)

Backend: Python (FastAPI), hosted on Replit

AI Model: Cohere Command R+

PDF Parsing: PyPDF

Deployment: Bolt for frontend, Replit CloudRun for backend

📁 Test Reports Used: We used publicly available annual reports including the 2024 Office of the Advocate for Small Business Capital Formation Report from official U.S. Government websites (like sec.gov) to validate our use case and build real-world relevance.

⚙️ Struggles We Faced:

💢 Frequent CORS and network issues when connecting backend to frontend

🧠 Cohere had token limitations, and we had to strategically extract only relevant chunks from PDFs

🐛 Handling large PDF files (300+ pages) often caused timeout and memory issues

🔁 Repeated backend crashes during demo testing, especially with heavy input

🔌 Learning how to integrate APIs across Bolt and Replit was a new but valuable experience

😵‍💫 Multiple frontend state bugs while tracking upload history and context switching

💡 Why We Built This: Reading and analyzing 200+ page reports is not only time-consuming but also inefficient for decision-makers. We wanted to bring clarity through AI, enabling people to quickly understand company performance, risks, and strategies in seconds rather than hours.

❤️ What Makes Us Proud: We’ve poured our soul, heart, and hundreds of debugging hours into Insightly AI. It's not just a project, but a meaningful step toward automating document intelligence.

And the best part? Our demo worked successfully — even on government PDFs!

📌 Try it | Share it | Fork it We’d love feedback from the community and professionals. Thank you to LTIMindtree for the inspiration and this opportunity.

AI #Cohere #FastAPI #Hackathon #PDFIntelligence #FinTechAI #StudentProject #MadeWithHeart #InsightlyAI #LTIMindtreeChallenge #ReactApp #BackendDev #Replit #AIInnovation

Share this project:

Updates