Private, offline academic tutoring — on any device.
Cleophis is a growing library of fine-tuned, on-device AI models — a full academic tutor that works 100% offline, with data that never leaves the device, sized to anything from a 4 GB phone to a 32 GB PC.
39 million US students are locked out of a $155 billion tutoring market — they can’t afford it, can’t get online, and their parents won’t trust the cloud with their data.
The trifecta of knowledge disparity: cost, connectivity, and privacy.
Cleophis answers all three.
A library of fine-tuned, on-device AI models — downloaded, not streamed — starting with a full academic tutor.
A full academic tutor for families priced out of $60/hour tutoring.
Works with no internet, on hardware from a 4 GB phone to a 32 GB PC.
Student data never leaves the phone or PC — nothing is logged to a cloud.
“Microsoft is exploring DeepSeek V4, a Chinese open-source AI model fine-tuned on Azure, as a cheaper model option for Copilot Cowork”
Axios, June 16, 2026 · Memeburn, June 19, 2026
The exact category Microsoft is now chasing — a fine-tuned, local model.
Zero setup, on-device execution.
The application profiles the device on launch, loads the best-fit model from its catalog, and runs 100% offline with zero cloud latency.
Checks the device’s memory (4 GB to 32 GB+)
Automatically downloads the best-fit local model for the device
24/7 tutoring, fully offline
From install to first answer in minutes — one sign-in, fully offline tutoring, and no student data leaving the device.
One on-device library. Any subject.
Academic tutoring is our wedge. The same architecture — small models fine-tuned to the subdomain and sized to the device — extends to anything people want to learn, privately and offline.
- Algebra II
- AP Biology
- Essay writing
- SAT · ACT
- MCAT
- Bar exam
- Coding
- Accounting
- Skilled trades
- French cuisine
- Guitar
- Home repair
The moat: not one general cloud model, but a growing library of narrow, on-device experts — each small enough to run on a phone.
Democratizing K–12 education.
We’re opening a $1.4 billion US market — starting with the 39 million students the cloud leaves behind. Cleophis is raising a pre-seed round.
Inference, the expensive part of AI, runs 100% on the customer’s device. Our costs — training, storage, a one-time download — are cents per user.
Additional usage adds no cloud or API cost, so margins hold steady as the user base grows.
In-house training and hosting of our model library — no reliance on frontier-model pricing or availability.