AI Engine
Multilingual content engine that verifies quotes instead of trusting AI memory
A multilingual content-intelligence engine that ingests lecture transcripts (Tamil, Arabic, English), extracts scholarly references, verifies them against a seeded source corpus, and indexes everything for cross-language semantic search, instead of trusting a model to recall scripture from memory. Async, queue-based pipeline: extract, verify against real source data, deduplicate into a shared registry, serve cross-language search. Currently powers a live public content platform (makthabahalqamah.com), built to generalize into future AI apps. Stack: Cloudflare Workers, D1, R2, Vectorize, Queues, Workers AI, Gemini. Had to re-architect the embedding approach after it underperformed on Tamil and Arabic content.
