gemma4-e2b

v0.dev, but open-source and local.

Better on-device macOS dictation: WhisperKit transcription, optional Gemma 4 post-processing.

Local document redaction, measured. A stock quantized Gemma E2B plus deterministic rails remove identifying spans on your own machine; nothing leaves it. Ships with the whole benchmark — 118 real documents, every miss named, and one offline command that recomputes every published number.

WeatherProof: An Agentic Vision Framework. Leveraging Gemma 4-E2B to intelligently mitigate snow, rain, fog, and low-light obstructions in real-time. From autonomous vehicles to critical security infrastructure—see clearly, anywhere, anytime.

AI-Powered Repair Assistant

Globis Edge: Offline caseworker companion [Powered by Gemma 4]

Local redaction for legal documents, measured. A frozen engine and a stock Gemma E2B replace the identities in an agreement, judgment or matter file with placeholders on your own machine; the blanked copy is what goes to a frontier model. Windows app, Apache-2.0, with the benchmark and its receipts.

An AI intelligence agent with Gemma 4, capable of handling files and folders

Colog – A fully offline AI-powered personal knowledge assistant that can index PDFs and YouTube videos, transcribe audio with Faster-Whisper, store embeddings in ChromaDB, and answer questions locally using Ollama + Gemma 3 4B.

AI medical assistant powered by a fine-tuned Gemma 4 E2B model with model (Q4/Q8) selection

FamilyVault - AI powered local end to end encrypted vault which uses power of gemma4 e2b to power vault.

AI support routing system with deterministic triage, bounded RAG, and local LLM inference for low-latency customer support.