Physics has not changed in the last 50 years. But computing architecture sure has.
Nullspace uses validated, reliable numerical methods to solve the same governing physics equations as the well-accepted simulation tools already embedded in R&D workflows across the industry.
With a few key differences:
Our solvers utilize a modern implementation of the Method of Moments algorithm.
They were developed for today’s GPU- and CPU-accelerated parallel computing architectures.
They were designed from the ground-up for Python-based AI-driven workflows.
Nullspace solves Maxwell's equations directly using a frequency-domain Method of Moments (MoM) solver with a Surface Integral Equation (SIE) formulation. Every result is a full-wave solution: no asymptotic shortcuts, no reduced-order approximations, and no AI surrogate predictions in the solver core. For novel hardware with no prior test data, you need a full-fidelity simulation you can trust from component- to system-level.
Our technology moat includes several intelligently designed computational advantages: adaptive matrix compression, which cuts the memory cost of large MoM matrices by orders of magnitude without any sacrifice of accuracy or result fidelity, and high-order geometry and high-order basis functions, which better capture the physics at lower computational cost than legacy solutions.
1
2
Nullspace does not run into either limitation
Our software delivers speed and accuracy during the design of system-integrated multi-band antennas, the RCS analysis of complex scattering bodies and the assessment of phased array radar system performance.
Because Nullspace stores a compressed representation instead of the fully dense impedance matrix, you can solve larger problems using the same computational hardware.
Where legacy tools stop, Nullspace keeps going
Phased array scaling, 15.5 GHz. A legacy tool exhausts memory above 8×8 (even on a 256 GB workstation) while Nullspace can simulate all 1,024 elements of a 32x32 array on one machine.
Nullspace was developed in stealth at an established U.S. defense contractor for 15+ years, and verified against real program requirements and validated on antennas, radars, and RF systems physically built and delivered to defense customers. Although that work isn't publicly releasable, it's the foundation our software was proven out long before Nullspace became a separate company in 2023 and launched Nullspace EM and ES as commercial software.
CMMC Level 2 Compliant.
Operable in air-gapped environments.
Currently deployed on multiple classified information systems.
Image credit www.rtx.com
Run Nullspace simulations natively from AI assistants
Because that interface is the language AI assistants were trained on, Claude, Copilot, and GPT can write and run Nullspace simulations natively, with no proprietary scripting layer in between.
AI removes the busywork, not the physics
The speed, scale, and rigor are properties of the Nullspace EM solver. AI doesn't make the physics faster or more accurate; it simply removes the busywork around the physics-based simulation.
No AI surrogate models in the solver
Other tools are beginning to predict electromagnetics with AI surrogate models trained on EM simulation data instead of rigorously solving the physics governing equations.
Trustworthy on novel designs with no prior data
That methodology only works in design spaces close to the data the model was trained on. It breaks in the most critical areas of the R&D design process: novel, next-generation hardware with no prior data to learn from.
Built for engineers, by engineers
How much are you leaving on the table?
The studies you skip, the designs you simplify, the deadlines you stretch. What if you didn't have to?