-
-
-
-
-
Code base of the Project
Inspiration
IPO Draft Red Herring Prospectuses (DRHPs) are among the most important documents for investors, yet they are rarely read end-to-end by retail participants. These documents often span 600–800 pages and are filled with legal, regulatory, and financial language that is difficult for non-experts to interpret.
As a result, many investors rely on YouTube summaries, social media discussions, or fragmented opinions instead of primary disclosures. This creates a major information gap and increases the risk of uninformed decisions.
IPO Lens was inspired by a simple question: What if investors could ask questions directly to the DRHP instead of searching the internet for summaries?
What it does
IPO Lens is an AI-powered research assistant that turns DRHPs into an interactive knowledge base. Users can upload an official DRHP PDF and: Ask investor-focused questions in natural language Automatically detect key risks and red flags Generate a simple one-page IPO brief Receive a transparent risk score based on disclosures Get grounded answers that avoid hallucination and clearly state when information is not disclosed The goal is to transform complex regulatory filings into accessible, investor-friendly insights.
How we built it
The project uses Retrieval-Augmented Generation (RAG) powered by the Gemini API.
Workflow: Extract text from uploaded DRHP PDFs Split the document into meaningful chunks Convert chunks into vector embeddings using Gemini Embeddings Store embeddings in a vector store for semantic search Retrieve relevant sections based on user queries Use Gemini models to generate grounded answers, summaries, and risk analysis The frontend was built using Streamlit to create a simple and interactive interface.
Challenges we ran into
Handling very large documents (hundreds of pages) efficiently Preventing AI hallucination when financial data is not disclosed Optimizing chunking and embeddings to work within free-tier limits Designing prompts that produce investor-friendly outputs rather than generic summaries Ensing stable embedding and API rate handling for large DRHP files These challenges helped refine both the architecture and prompt design.
Accomplishments that we're proud of
Built a working end-to-end RAG pipeline for real regulatory documents Achieved grounded answers that avoid guessing missing financial data Created a system that extracts investor-focused insights from complex filings Developed a functional prototype that can analyze any DRHP PDF
What we learned
Long-document understanding requires careful chunking and retrieval design Prompt engineering is critical for producing structured financial insights Avoiding hallucinations is essential in finance-related AI applications Gemini’s long-context and reasoning capabilities are powerful for regulatory documents Building AI tools for real-world problems requires balancing accuracy, cost, and usability
What's next for IPO Lens
Public hosting and real-time DRHP ingestion from SEBI Financial dashboards and visualizations (charts & metrics) Automatic comparison of multiple IPOs Page-level citations for every answer Support for annual reports and earnings filings Integration with fintech platforms and brokerage apps
IPO Lens aims to become an AI-powered research layer that makes capital markets more transparent and accessible.
Log in or sign up for Devpost to join the conversation.