AI Powered by Deskside Systems: Accelerated Compute Without the Cloud

AI applications have become synonymous with cloud computing, because the intense computation demanded by machine learning and inferencing could bring traditional workstations to their knees. But this doesn’t have to be the case, argued panelists in DE 24/7’s webcast titled “AI Powered by Deskside Systems,” April 16, 2026. In fact, with a professional engineering workstation and powerful new processors, AI-enhanced engineering, simulation and manufacturing workflows can run quickly and accurately at the engineer’s desk. 

“Humans can only use one application at a time, because they have to move the mouse and type,” said Sean Young, Director of Enterprise Industry Marketing, NVIDIA. “But Agentic AI can run multiple applications for you, as long as you have the compute capacity. That means, you need to have enough CPU and GPU cores.” 

Adam Smith, Head of Business Development, InfinitForm, described his firm’s use of Dell Precision systems with NVIDIA RTX GPUs to run Agentic AI-powered simulation. “We’re a physical AI platform. We use physics-based GPU solvers, deterministic algorithms, and an agentic AI pipeline. It takes engineering constraints as inputs to produce manufacturing-ready parametric CAD as output—not meshes, no manual reconstruction,” he said.

The knowledge from simulation software and the parameters from the CAD design are the starting points. “Parts that used to take weeks to design, can now be computed in minutes with InfinitForm and NVIDIA hardware,” he noted. One advantage of InfinitForm is, “You can run simulations for different manufacturing options.” 

Ken Flannigan, Industry Strategy and Alliances, Dell Technologies, pointed out, “The G in GPU stands for graphics, but don’t get caught up in the name. It’s an incredible parallel processor. If you have an engineering workstation, you can take advantage of the AI features available locally, if it’s properly configured.” And upgrading a system may be as simple as upgrading the undersized GPU.

“NVIDIA works closely with simulation software vendors to optimize for the GPU,” added Flannigan. “When you have a GPU-accelerated solver, the speed is greatly increased. You can do studies iteratively.”

Mark Burhop, Chief Technology Officer, AS-1.ai, also joined the panel to describe his software development with OpenClaw, a free and open-source AI agent that you can command using natural language. Aimed at CNC-based manufacturing, Burhop’s AI has to reconcile CAD parameters, safety regulations, physics-based simulation of the machine’s behaviors, and the pesky nature of manufacturing equipment. With so many parameters affecting the manufacturing outcome, Burhop developed a workflow he could control using OpenClaw. To add a security layer to OpenClaw so the pipeline is secure enough for enterprise deployment, Burhop recommended adding NemoClaw, developed by NVIDIA.

“[NemoClaw] puts things in a sandbox, so if something happens in your test, it limits the blast radius, as it were,” said Burhop. He relies on the Dell Pro Max workstations with NVIDIA GB10 processors. “These processors do the heavy lifting. They ingest Gigabytes of data from research and set it up for model building,” he said. “All my data is locally stored, so there’s no risk.”

You can watch the webinar on-demand here.