Why NavLive chose NVIDIA AI Jetson for AI scanner

Now you can train a 120B LLM at home using Unsloth and NVIDIA DGX Unsloth enables local fine-tuning of LLMs with up to 200B parameters on the NVIDIA DGX™ Spark. With 128 GB of unified memory, you can train massive models such as gpt-oss-120b and run or deploy inference directly on DGX Spark. NVIDIA DGX Spark is a workstation-class device for AI development and research. NVIDIA DGX Spark is powered by NVIDIA GB10 Grace Blackwell superchip which can deliver up to 1 petaFLOP of FP4 AI performance. It includes 128GB of coherent unified memory system, which avoids the overhead of moving data between VRAM and system memory. #llms #finetuning #nvidiadgx

Powering next-gen autonomous and embedded systems. The SBC3902, a compact rugged 3U VPX single board computer aligned to SOSA standards. Featuring the NVIDIA Jetson AGX Orin, it delivers up to 275 TOPS of AI performance with support for real-time GPU processing and deep learning algorithms. To learn more visit: https://lnkd.in/dXZHz42w #AMETEK #EmbeddedSystems #RuggedComputing

VersaLogic’s “Sabertooth AI” sets a new standard for embedded computing performance. It’s the all-in-one solution for developers working in space-constrained, high-performance environments. For example, you would need 25 of the NVIDIA Jetson AGX Orin to achieve the same amount of AI processing power that one Sabertooth AI embedded computer system can achieve. https://bit.ly/3VWhyiZ

Powering next-gen autonomous and embedded systems. The SBC3902 from Abaco Systems, a compact rugged 3U VPX single board computer aligned to SOSA standards. Featuring the NVIDIA Jetson AGX Orin, it delivers up to 275 TOPS of AI performance with support for real-time GPU processing and deep learning algorithms. To learn more, click the link below! https://lnkd.in/dXZHz42w #AMETEK #EmbeddedSystems #RuggedComputing

As transistor gains stall, hybrid bonding takes center stage. Dutch toolmaker Besi is turning precision and process lock-in into outsize margins—and a pivotal role in AI hardware. Besi Netherlands B.V. #advancedpackaging #leadingedge NVIDIA https://lnkd.in/eRbCYhPd

Today we unveiled HP's first offering in a brand-new category of devices purpose built for AI development: the ZGX Nano AI Station. Powered by NVIDIA G10 Grace Blackwell Superchip, this tiny powerhouse delivers supercomputer performance to run large AI models locally and accelerate time to results with the exclusive ZGX Toolkit. And yes—it fits in your backpack. Coming this fall. #ZGXNano #AIStation #EdgeAI #NVIDIA

Voxel51 is headed to GTC DC, Oct 28–29! Meet us at booth #411 for a live demo of the NVIDIA Omniverse NuRec + FiftyOne integration. As AV and robotics systems move from R&D to deployment, dataset quality remains one of the biggest challenges. Unvalidated data and fragmented workflows slow development, introduce costly errors into simulation pipelines, and drive up compute costs. Our integration with NVIDIA Omniverse and Cosmos addresses this challenge by creating a validated, enriched data pipeline ready for neural reconstruction and simulation. We’re powering the next generation of AV development alongside NVIDIA DRIVE. 📍See it live at #NVIDIAGTC DC and learn how this integration is redefining how synthetic data is built for Physical AI development: https://lnkd.in/eAKtcPqU 

The future of Physical AI is digital twin (Neural Reconstruction) and synthetic data generation but before you can even get there, your unwieldy data needs to be ready for it. Come see how Voxel51 and NVIDIA DRIVE have been working closely together to solve for this! Itai H Zadok Katie Burke Washabaugh Daniel Gural Matt Cragun Kirti Joshi Gautham Sholingar

I wouldn’t say no to having one. The world’s smallest #AI supercomputer, the NVIDIA #DGXSpark, just landed at SpaceX — hand-delivered by Jensen Huang himself to Elon Musk. For independent researchers like me, this is the kind of tool that could turn months of simulation into days. The #ΨΛ (Psi-Lambda) project already runs across multi-GPU setups, modeling logical-physical systems and hybrid spiking networks. A #DGXSpark would push it to the next scale — unified causal #training, real-time δΨ² / δ𝒮 metrics, and full-loop adaptive inference. And what a machine: • NVIDIA GB10 Grace Blackwell Superchip — up to 1 petaflop FP4 AI performance. • 128 GB unified CPU–GPU memory for seamless prototyping, tuning, and inference. • ConnectX networking + NVLink-C2C, giving 5× PCIe bandwidth. • NVMe storage and HDMI output for local deployment and visualization. • Full NVIDIA AI software stack pre-installed. Small but mighty indeed. If there’s ever room for a project exploring logical universes and neuromorphic unification, I’d gladly run it on Spark. #SparkSomethingBig

Engineering excellence on display: How three industry leaders solve the AI infrastructure equation. Cologix provides the 36MW carrier-dense facility with dedicated power infrastructure. Supermicro delivers air-cooled 8-GPU servers supporting rapid deployment timelines. Lambda enables 1-Click provisioning that scales from R&D to production seamlessly. Technical highlights that matter for enterprise decision-makers: → 286-pound servers with 6x 5,250W titanium-efficiency power supplies → 8 NVIDIA ConnectX-7 NICs per server driving 400Gb/s networking → Heat exchanger systems working like car radiators for efficient cooling → Tens of petabytes of NVMe storage ensuring GPUs never wait for data The video demonstrates why air-cooled deployment matters: faster time-to-market without complex liquid cooling infrastructure requirements. Watch the full tour from ServeTheHome : https://hubs.ly/Q03Jmfgc0 #AIEngineering #TechLeadership #InfrastructureInnovation #EnterpriseStrategy