Scholarly platforms
We build tools that help people find, read, and cite material accurately — from scholarly corpora to an organization's own knowledge base. The source stays as it is, and every answer points back to it.
What we build
Searchable corpora
Full-text search across large collections of Arabic texts, with handling for diacritics, letter variants, and search by root.
Bibliographic databases
Structured records of manuscripts, editions, authors, and works, with search, filters, and links between related records.
RAG over your collections
Retrieval-augmented question answering drawn only from your texts, with references to the exact passages used.
Enterprise knowledge bases
Connect the same approach to policies, manuals, wikis, and internal documents so staff ask in natural language and get answers grounded in your approved sources.
On your servers, not in the cloud
Many collections and knowledge bases cannot leave the institution, and many should not. We install the whole system, including the language models, on your own hardware.
Your texts and documents are never sent to an outside AI service unless you explicitly choose that.
- Runs on your servers or on a private server you control
- No data sent to third-party AI providers by default
- Can run on an internal network with no internet access, if required
- Integrates with the stores you already use for documents and knowledge
Reviewed by people who know the material
Software can retrieve a passage. It cannot judge whether an answer is sound. We involve subject specialists or business owners to review search behavior, test questions, and assess answers before a system goes live.
The system shows its sources, so readers can check every answer themselves.
Common questions
Can the AI answer from outside our collection?
Not unless you want it to. By default, answers come only from your texts and knowledge base, and the system tells you plainly when it cannot find an answer there.
Can you connect to our existing knowledge base?
Yes. We plan the connectors and indexing during scoping — file stores, wikis, and document systems you already run.
What hardware does it need?
That depends on the size of the collection and the number of users. We size the hardware during scoping and can work with what you already have.
Can you work with scanned books?
Yes, after text recognition. Arabic OCR needs review, so we plan correction into the work.
Who owns the platform?
You do. You receive the source code, the data, and the documentation.
Have a project?
Tell us what you need. We reply with questions, a rough scope, or a clear no if we are not the right fit.