OptiLearn v1.0.0 — Gemma 4 Good Hackathon Release
OptiLearn — Offline AI Classroom for Underserved Students
One teacher laptop. 30 students. No internet. Full AI tutoring.
OptiLearn turns a teacher's laptop into a local AI tutoring server. Students connect from any phone or tablet over a WiFi hotspot — no app install, no account, no internet required.
Built for the Gemma 4 Good Hackathon 2026 · Opti5 Labs
What's in this release
Web app (core product)
- Multilingual AI tutor chat powered by Gemma 4 via Ollama — runs fully offline
- Live class mode: teacher speaks → Whisper transcribes → AI translates to every student's language in real time
- Translate & Learn: upload a PDF or photo, get instant translation + grade-level explanation
- Adaptive quizzes with EMA mastery tracking per student per topic
- Text-to-speech in 14 languages (Piper) + 60+ languages (MMS-TTS) — including Somali, Amharic, Tamil, Urdu
- Mastery heatmap, automatic alerts, quiz builder, material upload, PDF reports
- Kahoot-style live quiz game
- Offline-first PWA — installable on any Android or iOS device
- Online mode: transparent fallback to Gemma 4 26B / Gemini API when internet is available; student chat always stays local
Desktop app (bonus installer)
- Electron wrapper — no Python, Node, or terminal required
- First-launch setup wizard: API keys → Ollama install check → AI model download with progress bar
- macOS DMG (Apple Silicon) and Windows NSIS installer (x64 + ARM64)
Fine-tuning pipeline
- Unsloth QLoRA pipeline for Gemma 4 on trauma-aware multilingual pedagogy
- 11,500+ training examples across Arabic, Somali, Amharic, French, Swahili, Tigrinya, Hausa
- Runs on Kaggle free tier (A100, ~3.5 hours)
Getting started
Web app:
git clone https://github.com/Ilakiancs/OpitLearn
cd OpitLearn/optilearn
./scripts/setup.sh
cd frontend && npm install && npm run build && cd ..
./scripts/start.shDesktop app: download the installer below for your platform.
See the README for full setup, offline classroom configuration, and troubleshooting.
Requirements (web app)
- Python 3.11 or 3.12
- Node.js 18+
- Ollama with
gemma4:e2bpulled - 8 GB RAM minimum (16 GB recommended)
- No GPU required