Simulated Verified Inference Pipeline — interactive demo

Cyberian Systems' verified-inference platform turns any AI inference call into a tamper-evident, independently-verifiable record. This page renders an in-browser simulation of the four stages every job moves through:

1. Submit — the client sends inputs to the coordinator over HTTPS, authenticated with a Bearer API key.
2. Compute — a deterministic worker executes the requested model (e.g. BGE-large embeddings, ONNX, CPU-deterministic) inside an isolated, attestation-capable sandbox.
3. Verify — independent verifiers re-execute the computation and check that the outputs match the worker's, byte-for-byte.
4. Receipt — a cryptographically signed Merkle receipt is emitted, anchoring inputs, outputs, and the verification attestation in tamper-evident form. The receipt can be re-checked offline at any time via client.verify_receipt().

The simulation here runs entirely in your browser — no API call is made and no live inference is performed. For the production version that talks to the live coordinator, see live-demo.html. The full SDK documentation, including authentication, models, and receipt verification, lives at docs.html.