Sovereign scientific computing · a real kernel on your own machine
Drop a dataset, ask a question in plain language, and get a real, reproducible analysis — run by a full Python kernel on this machine: every package (scikit-learn, SciPy, PyTorch…), GPU-ready, big files, no browser sandbox limits. Your data never leaves the device. Built on EU rails, GDPR-native.
Don't take our word for it. Pull your network cable or turn off Wi-Fi and keep working — the analysis still runs, because the Python kernel is on this machine. Or open your browser's Network panel (or a firewall like Little Snitch): the only traffic is 127.0.0.1 (this machine) — never your dataset.
The kernel and client are small and open to inspection — have your security team read exactly what runs. Full safety brief →
1 Security level
Pick it before adding data — it sets what (if anything) ever leaves your device. Labelled to match TU Delft's own data classification: Critical / Sensitive / Standard.
2 Dataset
What actually happens: your CSV is read, analysed and charted by a real Python kernel running on this machine — the same environment you already have (pandas · NumPy · Matplotlib · scikit-learn · SciPy · anything on PyPI · your GPU). The browser tab only talks to 127.0.0.1; nothing is uploaded. Every cell's code is shown, so results are reproducible and auditable — the way a notebook should be. This is what the web version can't do: no WebAssembly ceiling, real files, real compute — and the data still never leaves the device.
prototype reference implementation of the desktop notebook — the local kernel (kernel.py) serves this page and runs your code. In the packaged MakeMode app the same contract is served by the Tauri control server, so the notebook lives inside the app window.
· All-Scaleway, France · No US infrastructure ·
how sovereignty works →