Company3 min read
RadMah joins NVIDIA Inception to generate data you can prove
RadMah, Inc., the company behind RadMah AI, has been accepted into the NVIDIA Inception program, and plans to evaluate accelerated compute for an engine whose every run seals its own evidence.
On 14 September 2026, RadMah, Inc. was accepted into the NVIDIA Inception program, and we are pleased to be a member.
We built RadMah because the most useful data in any organisation is usually the data you are not allowed to touch. RadMah AI generates realistic data for AI, software testing, healthcare and industrial systems. Fabricate takes a plain-English description and returns a validated dataset. Synthesize learns the shape of your own data, reproduces it without copying rows, and scores the result against your source. Healthcare FHIR produces HL7 FHIR R4 bundles without source patient records.
You cannot attack a live plant to learn what an attack looks like, so our industrial engines simulate one. Virtual SCADA produces plant telemetry from physics-based process models. Virtual PLC is a controller simulator you run inside your own network, and its Enterprise edition exposes five live protocol surfaces: Modbus TCP, OPC-UA, BACnet/IP, MQTT and DNP3. ICS Security produces labelled attack scenarios, and where we lack a faithful generator the engine refuses rather than emitting an unsupported label.
What “you can prove” means
Realistic is not enough. Every run seals a hashed, tamper-evident evidence bundle: the sealed contract, the constraint report, an index of every artifact with its own BLAKE3 hash, and the determinism report. The same sealed contract, the same seed and the same engine version reproduce byte-identical data. You do not need us to check it: rady evidence verify, shipped with the SDK and CLI, runs offline with no API key and no call back to RadMah.
“Generating something plausible was never the hard part. The hard part is handing someone a dataset and a bundle they can verify on their own machine, without trusting us. That is what RadMah is for, and membership arrives just as we want to push the engine harder while that promise holds.”
What we plan to explore
Our synthesis engine runs on CPU, CUDA-capable NVIDIA GPUs and Apple Silicon, with deterministic algorithms pinned on every backend; our production fleet runs on CPU today. Through Inception we gain access to the latest NVIDIA SDKs, model libraries and developer platforms, the NVIDIA Developer Forums, free self-paced NVIDIA Deep Learning Institute courses and discounted expert-led workshops, and technology guidance and connections within NVIDIA’s ecosystem. Preferred pricing on select NVIDIA hardware and software, and cloud credits from NVIDIA and partners, are available to request; we have requested none yet.
We plan to evaluate accelerated compute in three places: training and sampling in the tabular synthesis core, where longer epoch budgets and wider sweeps in the same wall-clock are how fidelity on long-tailed data is won; the quality battery that gates every job, so an engineer gets a verdict sooner; and how many simulated plants we can hold live at once, which caps ICS dataset size.
“Our promise is proof, not speed for its own sake. Any faster execution path has to return the same bytes under the same contract and seed, or it does not reach a customer job. If it passes, customers wait less for the same evidence bundle; if not, it stays out of the product. That rule does not bend for a GPU.”
What comes next
Measured evaluation first: run the same sealed contracts on accelerated hardware, compare the bytes, and ship nothing until the determinism report agrees. We will report what we learn here on radmah.ai. RadMah AI is available today as a managed service in the United States or the European Union, or self-hosted as an offline container bundle.
About RadMah
RadMah, Inc. is a Delaware corporation and a member of the ITLOX group of companies. It operates RadMah AI, a synthetic-data and simulation platform that generates realistic data for AI, software testing, healthcare and industrial systems, with every run validated and sealed into evidence of what was produced and how.
NVIDIA, the NVIDIA logo and NVIDIA Inception are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries. RadMah, Inc. is a member of the NVIDIA Inception program; membership does not constitute an endorsement, partnership or investment by NVIDIA.