NumericalAI is developed and operated by Empirisch Tech GmbH, a company registered in Vienna, Austria (empirischtech.at). The platform is hosted under the numericalai domain (numericalai.net), which is Empirisch Tech's brand for AI and engineering solutions.

Unlike traditional simulation tools, NumericalAI is built using advanced GPU acceleration and AI-driven algorithms. This combination drastically reduces simulation time, minimizes errors, and enhances efficiency and productivity.

When benchmarked against a 32-core CPU system, NumericalAI achieves up to 10× faster performance on common engineering workloads like CFD simulations. It was observed that the speedup factor increases with the complexity of the simulation. This is made possible through GPU parallelization, AI-optimized pipelines, and intelligent resource allocation.

The folder must contain a config.json file and any other supported files to run the simulation. The file must be exactly named 'config.json' and not a folder with the same name.

This is the Palace simulation configuration file — a JSON file that defines the input parameters for your simulation. Detailed explanations and guidance on how to complete it are provided in the user manual.

New users activate their account with the €1 Starter plan, which includes two hours of compute time and basic GPU access. Purchasing Standard, Professional, or additional compute credits upgrades you to On Demand Usage (Premium), with priority queue access, advanced AI-assisted validation, and debugging tools.

The Enterprise plan offers unlimited compute time with the GPUs of choice (H100s, A100s, etc.) and a dedicated queue, allowing unlimited parallel simulations without waiting. It also provides access to advanced capabilities like Structural, Multiphysics, and Molecular Dynamics simulations, with optimized, cost-effective solutions tailored to your specific needs.

Directly click on the 'Contact Sales' button placed under the Enterprise plan section on the Buy Credits page and we will get back to you shortly. Otherwise, you can also use the contact form under Settings page to write a custome message and request a quote.

If your simulation's estimated runtime exceeds 25 days, please contact us to explore a custom solution. Simulations running longer than 29 days are automatically cancelled.

Got to settings page and under 'Manage Billing' you can find all your invoices.

If you are registed as a business customer, you can get your VAT back by providing your company name, address, and valid VAT ID thorugh the contact form under Settings. Please ensure your information is complete and accurate at checkout. We will then issue a separate invoice with the VAT deducted.

Our compute fleet currently consists of NVIDIA A100 and H100 GPUs, offering exceptional performance and scalability for large-scale engineering and scientific simulations.

Processing time includes additional overhead such as job initialization, environment setup, data transfer, and post-processing. The Palace logs only show the core simulation runtime, not these extra steps.

You can run GPU-accelerated computational electromagnetics (CEM) simulations with Palace, AWS's open-source 3D finite-element solver. Supported analyses include eigenmode, frequency-domain, and time-domain electromagnetic problems — such as resonators, S-parameters, antennas, signal integrity, and quantum hardware — using your Palace config.json and mesh files. Workloads run on NVIDIA A100 and H100 GPUs for fast, scalable performance.

€1 Starter users can run or queue only one job at a time, while Premium users can run and queue up to three jobs simultaneously, benefiting from faster access and reduced waiting times.

The estimated time is calculated based on the average time per step of the simulation and is updated on every refresh. The ETA is highly accurate as it only measures the actual simulation time. After simulaiotn stage, there is additional steps such as post-processing, data transfer, compression, etc., which could last longer (e.g., an hour or so for simualtions generating several GBs of data) depending upon the size of the simulation results.

Please use the Contact Us form available on the Settings page for assistance. Our support team typically responds within 24 hours on business days.

NumericalAI uses state-of-the-art solvers and validated numerical models to ensure high accuracy and numerical stability. Results are benchmarked against analytical and experimental data where applicable.