Hi DEV 👋 I'm Alessio, CEO of Swiss Group Network, the Swiss company behind NEURAX Compute. We are launching on 31 December 2026 and we'd love feedback from developers before that.

The problem

Cloud computing is easy to buy and hard to trust. You send a job, you get a number back, and you pay. You can't easily check where it ran, who ran it, or whether the result is actually correct.

What NEURAX does

NEURAX Compute runs compute jobs on a distributed network of nodes, and exposes everything through a plain HTTPS API:

Verified results: every result is checked before you are charged. A node that returns a wrong result is not paid, and the work is reassigned.
You choose where it runs: per job, you can keep the work in Europe or in Switzerland (region), and limit it to identity-verified partners only (sovereign).
Signed evidence: every job comes with a digitally signed evidence file. Anyone can verify it, even without an account.
Paid in euro: no tokens, no crypto, double-entry accounting.

Workloads available today include portfolio risk (parametric and historical VaR / Expected Shortfall, Bitcoin included), Monte Carlo option pricing, matrix multiplication, containerised programs and 3D rendering.

What it looks like in Python

The official client has no dependencies:

bash
pip install "git+https://github.com/neurax-compute/developers.git#subdirectory=python"
export NEURAX_KEY="nxk_your_key"
python
from neurax import Client

nx = Client()
job = nx.submit("hist-var@1", params, region="EUROPE")

r = nx.wait(job["id"]) # checked before you are charged
print(r["var_99_eur"], r["es_99_eur"])

nx.evidence(job["id"], "proof.json") # signed proof, verifiable by anyone

You can verify an evidence file in the browser at https://neuraxcopute.it/verifica, or from code with nx.verify_evidence(...).

For developers

Everything you need is public on GitHub:

👉 https://github.com/neurax-compute/developers

OpenAPI 3.1 specification, to generate a client in any language
Python client (MIT license)
Ready-to-run examples in curl, Python, JavaScript and Excel (Power Query)
Guides on risk with Bitcoin, evidence files, containers and rendering

Full documentation: https://neuraxcopute.it/sviluppatori

To be transparent: the client library and the examples are open (MIT), while the compute engine and the coordinator are proprietary.

We'd love your feedback
Is the API shape clear? What would you change?
Which workloads would you want to run on a verifiable network?
What would make you trust a result from a distributed network?

Questions and ideas are welcome in the comments here or in the Discussions tab of the repository. Thanks for reading!