FAQ
Questions, answered.
Everything you might want to know about certifying robot policies with Certisto — from how certificates work to security and pricing.
Platform basics
Certisto is a Physical AI certification platform. It builds a physics-accurate digital twin of your robot's environment, validates policies at GPU scale, and issues an auditable Real-World Readiness Certificate predicting field performance.
Robotics teams deploying learning-based policies — humanoid startups, industrial-arm OEMs, warehouse automation, and autonomous-machine makers — plus the safety, compliance, and leadership stakeholders who sign off on deployments.
Simulators tell you what happened in a model. Certisto adds a calibrated sim-to-real estimator, a signed certificate artifact, and a telemetry flywheel — turning simulation into defensible readiness evidence.
Certisto runs on GPU infrastructure. We offer a managed cloud so you don't need your own hardware, and Enterprise supports on-prem and DGX deployment if you do.
Certificates
A calibrated readiness score, confidence bounds, ranked failure modes, a certified operating envelope, safety-violation counts, and full provenance (twin version, seeds, and signature).
Accuracy depends on twin fidelity and flywheel data volume. Across current customers, certificates correlate around 94% with field outcomes; every pilot includes a correlation review to validate this for your case.
Yes. Each certificate is cryptographically signed and reproducible from the record stored in the Evidence Vault.
Re-certifying a new policy version issues a new certificate. Regression tracking compares versions so you catch readiness drops before release.
Twins & data
Twin Studio authors OpenUSD environments with physics-accurate materials, lighting, and calibrated sensor models, then scores fidelity against your real hardware.
Data Foundry combines domain randomization with world-model synthesis, and the Counterfactual Scenario Miner actively searches for the rare cases that change a verdict.
No. Your twins and policies are tenant-isolated. Flywheel improvements use measured sim-to-real deltas under terms you control.
Pricing
A predictable platform subscription plus metered GPU usage, with certification priced per validated policy or robot SKU. Pilots start at a fixed POC fee; see the pricing page for details.
Yes — most teams begin with a paid Pilot that certifies a single policy, then move to Team or Enterprise.
We offer startup and research pricing for early-stage robotics teams. Ask during your demo.
Security
Data is encrypted in transit and at rest, isolated per tenant, and you can deploy fully on-prem or on your DGX so sensitive IP never leaves your infrastructure.
We build to SOC 2 and ISO 27001-aligned controls and handle data per GDPR and CCPA. Ask for current certification status during security review.
Yes. Enterprise supports running the entire suite inside your network with the same certificate format as our cloud.
Integrations
Natively with the NVIDIA robotics stack (Omniverse, Isaac, Cosmos, TensorRT, Jetson) and openly with MLOps, CI/CD, data, and identity tools via connectors, a REST API, and the certi CLI.
Yes. Trigger validation from GitHub Actions, GitLab CI, or Jenkins and block merges that fall below your minimum certified score.
Yes — a documented REST API with webhooks and client libraries for Python and TypeScript, plus a CLI for local and scripted runs.
Still have questions?
Talk to a real solutions engineer.
If your question isn't here, we're happy to answer it directly.
Ready to see a certificate on your robot?
Book a technical demo and we'll validate one of your policies against a twin of your environment.