RACHEL AI: Offline Semantic Search
Help us bring AI-powered search to offline communities worldwide
Prototype demo — semantic search across multiple RACHEL modules
The Problem
Right now, search on RACHEL is basically ctrl+F. Keyword matching within individual modules — and not even all of them support it. No way to search across the whole library.
The Solution
Students type a natural question like "how do plants make food" and get relevant results from across all their content — ZIMs, HTML modules, Kolibri channels, user-uploaded materials — grouped by source, ranked by relevance.
It understands meaning, not just keywords. Runs entirely on the device, no internet needed.
What $50,000 Covers
- GPU Compute (~$8–10k): Each module needs AI processing. A single large module takes 18+ hours on a GPU. Over 100 modules total.
- Storage & Infra (~$5k): Indexes are large. Need staging to build, test, and package for device deployment.
- Engineering (~$30–35k): Automated indexing pipeline, size optimization, parsers for every content format, deploy tooling, QA.
- Contingency (~$2–3k): AWS for periodic re-indexing, monitoring, bug fixes.
Timeline
Goal: Fully funded by May 31, 2026
📋 Donor Advised Funds (DAFs): We accept DAF contributions. If you donate via DAF, please let us know so we can update the tracker. Reference "RACHEL AI" in your grant recommendation.
World Possible is a registered 501(c)(3) nonprofit organization. Your donation is tax-deductible to the extent allowed by law. EIN: 26-4035658