Prajñā
A knowledge workspace that connects practical guidance with answers grounded in source material.
Public platform · kb.advayaglobal.com 
A problem worth working on.
Operational knowledge is only useful when people can find the right answer, understand its context, and get back to work.
From intention to implementation.
- Built a searchable knowledge library with document import and country-aware retrieval.
- Combined keyword search with optional semantic matching to find relevant passages.
- Built Ask Maharishi to return source passages by default, with optional semantic matching and an experimental local-model mode.
Where intelligence fits.
The default Maharishi assistant combines MiniSearch, phrase/concept matching, country filters, and short conversation context to retrieve verbatim source passages. Optional Transformers.js/MiniLM embeddings improve semantic matching in a browser worker. An experimental WebLLM/Qwen3-0.6B mode can explain retrieved context on compatible WebGPU hardware; the default assistant needs neither a model download nor a paid API.
What exists today.
A working library and question-answering flow, with a source-first fallback for devices that cannot run a local language model.
The optional generative mode requires compatible WebGPU hardware and a substantial model download. Physical-GPU quality validation is still an experimental boundary. Public retrieval uses public content; authorized private material follows access rules.
Technical descriptions are based on a review of the project’s source and available endpoints in October 2026. They describe implemented capabilities, not independently measured business outcomes.