Computer vision & automation
that stays on your premises.
We turn your existing cameras and systems into real-time intelligence — PPE and safety checks, retail analytics, and RPA — running locally, so sensitive data never leaves your site.
See what your cameras have been missing
One small edge appliance connects to your IP cameras and runs models on site. Get instant alerts and dashboards — for compliance, safety and operations.
PPE & compliance vision
Automatically verify that every worker on the line wears the required protective equipment — hats, hairnets, gowns, gloves, masks — in food, pharma and electronics plants.
- Hat & gown / hairnet detection
- Glove, mask and apron checks
- Restricted-zone & line-entry alerts
- Shift compliance reports
Store intelligence & hotspots
Turn existing cameras into a sensor network: count visitors, map heat-zones and dwell time, measure queues and conversion — without facial recognition.
- People counting & footfall
- Hotspot / dwell heatmaps
- Queue & wait-time alerts
- Shelf & zone engagement
Safety & process monitoring
Detect missing equipment, unsafe behaviour, blocked exits, forklift proximity and process deviations — and trigger an instant local alert.
- Safety & PPE monitoring
- Forklift / pedestrian proximity
- Blocked-exit & spill detection
- Cycle-time & anomaly flags
100% on-premise. Your data never leaves the building.
Every model runs on local edge hardware we install at your site. Video is processed locally and discarded or stored under your control — nothing is sent to external servers. That means lower latency, lower bandwidth cost, and a far smaller compliance and privacy footprint.
Don't take our word for it. Test the claim right here.
Pick a model below and it downloads straight into your browser, then runs on your own GPU. What you type and what it answers are never sent to any server — not even ours. Everything happens inside your device.
Once the model has loaded, turn off your internet connection and see for yourself: it keeps working, because your data never leaves your device.
Best with a WebGPU browser (Chrome / Edge 113+, Safari 18+, recent Firefox); without it the model falls back to the CPU, much more slowly. The model is downloaded straight from Hugging Face — it never passes through our server.
How does it work?
WebGPU and Transformers.js run an open language model (Gemma 4, Llama 3.2) directly in your browser. The model weights are downloaded once from Hugging Face and stored in your browser cache; every answer after that is computed on your own GPU. Our server does nothing but serve the HTML of this page.
Is nothing really sent anywhere?
Correct — and you can verify it yourself. Open the Network tab in DevTools and keep it open while you chat. The only traffic you will see is the model download from huggingface.co; there is no request carrying your text. For an even simpler proof: once the model has loaded, switch to airplane mode or unplug the network cable. The chat keeps answering. Nothing could be leaving your device.
What do I need?
Ideally a WebGPU-capable browser (Chrome / Edge 113+, Safari 18+, recent Firefox on Windows) and enough free memory for the model you pick. Without WebGPU it still runs, on the CPU. The smallest option is a ~1 GB download and runs on ordinary laptops. Gemma 4 is a 3-5 GB download and noticeably smarter, but needs a real GPU.
How do I delete the model afterwards?
It lives in your browser cache, so the button below removes it. Clearing site data from your browser settings does the same. If you come back later, the model will simply download again.
Is this the same as your AI services?
No — this is a technology demo. What runs here is a general-purpose open model; it was not trained on LoreaTec data. Real deployments run larger models on edge devices or servers installed at your own site, combined with your operational data. The principle is identical: the data stays in the building.
Robotic Process Automation (RPA)
We automate repetitive work two complementary ways — and combine them when it pays off.
Script-based automation
Reliable, auditable bots for structured, rules-based tasks — fast, predictable and inexpensive to run.
- Data entry, reconciliation & migrations
- File, email and report generation
- System-to-system integration without APIs
- Scheduled, logged, repeatable runs
AI-assisted automation
For unstructured inputs and judgement — LLM-powered steps that read, summarise, classify and draft, with humans in the loop where it matters.
- Document understanding & extraction
- Email triage, classification & replies
- Summaries, translation & knowledge search
- Deployable on-premise for sensitive data
Our own on-premise AI platform
Our AI vision & automation runs on the LoreaTec platform, deployed on edge hardware installed at your site — nothing is sent to external servers. Tell us your use case and we'll scope a pilot.
