May 20
Loading video player...
Full automation pipeline that scores companies on ESG (Environmental, Social, Governance) disclosures automatically, in minutes, instead of days of manual analyst work. What it does: Upload an Excel file with company names, PDF report links, and a scoring rubric. The tool downloads each sustainability report, extracts relevant text using AI-powered vector search, scores every topic with GPT-4o (GPT-4o-mini used for tests), verifies every quote against the source document to catch hallucinations, and delivers a colour-coded Excel results file to Google Drive with an email notification when done. Tech stack: Python FastAPI: REST API with async-safe background job processing OpenAI Responses API: structured outputs with Pydantic validation FAISS vector search: RAG pipeline for relevant passage retrieval PyMuPDF: layout-aware PDF text extraction rapidfuzz: hallucination detection via fuzzy quote matching n8n: workflow automation (webhook trigger, polling loop, Google Drive upload, Gmail notification) ngrok: secure tunnel to expose the local API to n8n Cloud Production-grade features: AI kill switch, rate limiter, concurrent job semaphore, atomic file writes, and a quote verification audit column in every output file.