Overview
Engineering contributor on a 4-person 5CVC team building tooling for Alpha Edison. Shipped a Chrome extension + RAG chatbot now in active use by the investment team. Cut diligence prep ~70%, saving associates 10+ hours per week. First AI tool I deployed in a professional VC context, not a toy project.
How it worked
The Chrome extension injects a content script into the active tab via Chrome's scripting API, scrolls programmatically to load infinite-scroll sections, then extracts visible text and the URL. Mercury Parser converts unstructured pages into structured readable articles, removing ads and navigation. Output goes to Retool through a secure API request with title, URL, timestamp, and scraped content. A RAG chatbot then lets analysts query that content conversationally, keeping the LLM grounded in actual scraped documents instead of hallucinations.
- Browser extension, not a separate tool. Lowest friction way to capture data during the analyst's actual browsing workflow.
- Mercury Parser for clean readable content. Tesseract OCR as a fallback layer for non-parsable sites.
- Retool stood up the live internal tool fast without custom infrastructure.
- Stack: JavaScript (Chrome Extension), Chrome Scripting API, Mercury Parser, Retool (SQL + workflows), Tesseract OCR, OpenAI APIs.