The Steele Foundation

1. Built a retrieval-augmented classification pipeline that transforms unstructured text into structured recommendations. The system embeds each input, retrieves the most relevant matches from a curated vector index, and passes those candidates to a GPT-based classifier to recommend standardized categories, including Canada's National Occupational Classification (NOC) codes.4:02
2. 02.Developed a hybrid rules engine that paired deterministic checks with LLM-based reasoning to flag key requirements and policy-sensitive conditions in unstructured documents.3:18
3. 03.Tested and improved Python/FastAPI services and Azure workflows that took documents from ingestion through AI analysis to human-in-the-loop review.3:55

℗ 2026 Sydney Gahunia · The Steele Foundation