Contact
Build with us. Support the project.
Tell us about your scientific ML workload or your organization’s interest in sustaining omnibias. We will discuss a scoped evaluation, integration, or corporate sponsorship.
Derivative-acceleration audit
We benchmark a closed-form replacement for your hottest operator against your current autodiff path - on your problem sizes.
Certified-numerics report
We assess whether a model statement fits a supported enclosure or certificate method, then document its assumptions, bounds, and verification status.
Integration project
We help integrate supported derivative paths into your PyTorch or JAX pipeline and validate accuracy, shapes, and parameter gradients.