Offline AI for K12
Put secure AI in every classroom
regardless of zip code, budget, or bandwidth.
We’re grateful to the
Offline Internet Consortium for providing funding for teacher stipends.
Click here to join the experiment.
Cloud-based, corporate, and opaque.
Teachers and students consume it without ever understanding it
The Problem
The Solution
Strip away the cloud.
Run open-source AI on hardware you physically own.
Nothing leaves the room
Offline AI for K12:
Hardware
- Raspberry Pi 5″
- The low-cost, single-board engine of the system
LMS - MoodleBox
A complete
learning management system,
entirely offline.
AI MODEL — Gemma 3:1B
Google’s lightweight model
supporting over 140 languages
COST — Under $200
One-time,
per classroom.
No recurring fees
Designed for educators, not engineers
No coding required.
This is for the curious, the skeptical, and the innovative, regardless of tech background
- Teachers, Grades 4–8, core instructional staff implementing AI-enhanced activities in everyday classrooms
- Media specialists & librarians who support and extend AI literacy across the building
- Instructional coaches guiding teacher adoption and workflow design.
The Deliverables — Concrete things, not just a certificate.
A portable AI environment
- MoodleBox running on your own Raspberry Pi 5
Hands-on LLM experience
- Prompting and evaluating Gemma 3:1B, fully offline.
A professional portfolio
- Teacher workflows and student AI-literacy lessons, ready to use.
The Four ideas behind the program
- Build understanding first. — Teachers who understand AI make better decisions about when — and when not — to use it with students.
- An offline environment empowers. — Teachers and students experiment more boldly when nothing can go wrong externally.
- It’s yours. — You have a playground, not a prescription. You decide how AI fits your specific classroom.
- Investigators, not consumers. — Students become critical investigators, not passive consumers. Teachers become builders.
Course Outline — Six Sessions
The Professional Development Course — Six sessions, roughly 90 minutes each. Asynchronous and self-paced throughout.
- Session 1 · What Is AI — Compare cloud AI to local MoodleBox AI. Opportunities, constraints, and what each reveals about how AI actually works. Grounds teachers in AI concepts before any hardware work.
- Session 2 · Building the Offline Environment — Boot the Pi, fire up MoodleBox, and talk directly to Gemma 3:1B, all without touching the internet. Hands-on hardware: unbox and configure your own Pi 5 + MoodleBox.
- Session 3 · Prompting & Hallucination Lab — Intentionally break the model: provoke false citations, biased answers, and overconfident errors, then build a shared evaluation rubric. The model is pushed to fail on purpose.
- Session 4 · Designing Teacher-Facing Uses — Draft rubrics, differentiate instruction, and brainstorm lesson hooks using local AI without risking student privacy. Shifts from tinkering to practical, real classroom workflows.
- Session 5 · Teaching Students About AI — Students roleplay as Bias Investigators and Fact-Checking Teams, using the Experience AI framework adapted for your classroom. Turns teachers’ understanding of AI failure modes into student lessons.
- Session 6 · Portfolio Showcase & Next Steps — Present your functioning hardware setup, share your mini-unit, and map out a responsible AI implementation plan for your school. The capstone, including peer feedback and a concrete next-steps plan.
Let's talk about how this fits your school
Sign up here
We’re grateful to the Offline Internet Consortium for providing funding for teacher stipends.
Contact Dan McGuire, dan@sabier.org, for more information.