Description
What if your AI automation never had to touch the cloud?
In this demo, I show how UiPath can trigger a fully local, offline AI model — Gemma 3 (4B), running through Ollama — with zero cloud API calls and zero data leaving the machine during inference. I prove it live by disconnecting Wi-Fi mid-demo and showing the model still works.
What you’ll see:
Gemma 3 (4B parameters) running locally via Ollama
Live proof of offline execution — Wi-Fi disconnected on camera
A UiPath automation (Macbook_EdgeAI) triggering the local model
End-to-end execution from UiPath Assistant, with real-time input
Orchestrator job logs confirming the local model call
Why this matters:
Most AI automation relies on cloud-hosted LLMs — great for flexibility, but not always viable for financial transactions, healthcare records, or any process involving confidential data. This demo shows a practical pattern for keeping sensitive data fully on-device, while still getting UiPath’s orchestration, governance, and Orchestrator visibility.
Tech stack:
UiPath Studio Web (workflow design)
UiPath Orchestrator (job management, logging)
UiPath Assistant (trigger)
Ollama (local model runtime)
Gemma 3, 4B parameters (Google’s open-source LLM)
Link
Date
2026-07-26
Related UiPath products
Agentic Process
Agents
Automation Cloud
Cloud Robots
Studio Web