The problem
When Muslims have a question about an Islamic ruling, the internet gives them a mess. Forum posts from unverified sources, opinions without attribution, and content from places nobody can trace back to actual scholarship. The knowledge exists in the books of established scholars, written and verified over centuries. But those books are not searchable the way people ask questions today. That gap is what Ilm AI was built to close.
The approach
- Built a RAG system grounded entirely in verified Islamic scholarly texts
- The ingestion pipeline processes content from established fiqh sources with careful attention to preserving the context and ruling structure of each passage, since breaking a ruling mid-sentence can completely change its meaning
- The retrieval step finds the most relevant passages before generating a response, and the LLM layer is constrained to answer only from what was retrieved, never from its general training data
- Every response shows the source, so users can check the scholarship themselves rather than just trusting the output — MVP shipped and live on GitHub
The result
Islamic rulings searchable in plain English, with every answer traceable to a specific scholarly source. No hallucinated answers, no random internet opinions dressed up as rulings. A tool that only works if the retrieval is genuinely trustworthy, and it is.