[Blog] Generative AI is moving from "answering" to "acting." Starting with agent-type AI using NVIDIA Nemotron.

AI Development, Local LLM, Deep Learning AI Workstation

The Applied AI Workstation refers to a high-performance computer optimized for advanced AI development tasks that require vast computational resources, such as deep learning, machine learning, data analysis, and simulation. By equipping the latest generation CPUs, large-capacity memory, high-speed storage, and high-end GPUs, it enables fast and stable execution of training and inference processes for large-scale neural networks. Additionally, with a software environment and development tools optimized for Linux-based operating systems like Ubuntu, it provides flexibility and scalability that allows researchers and developers to start working immediately. It can accommodate a wide range of industries and applications, including not only AI development but also scientific computing, image processing, and CAD/CAE analysis. Hardware configuration for AI development, local LLM, and deep learning workstation.

• Desktop PC
• Image analysis
• Image Processing

Rack-mounted servers for HPC, LLM, and generative AI development.

High-spec CPU models required for HPC High-end GPU models necessary for LLM/generative AI development We will propose the optimal models tailored to your needs based on usage, size, specifications, and price, ranging from 1U to 4U.

• Thermo-fluid analysis
• Desktop PC
• Other research software

Introducing models for LLM (Large Language Model) training and models for inference!

For LLM (Large Language Model) training purposes, here it is! APPLIED HPC Deep Type-AS4UX2S8GP-AU019 ■ CPU: Xeon Gold 6530 (32 cores/64 threads/2.1GHz/tb4.0GHz) ■ Memory: 2,048GB (64GB×32) ■ Storage: 1.92TB SSD + 7.68TB SSD ■ GPU: NVIDIA H100 94GB ■ OS: Ubuntu 22.04 LTS ■ Framework: TensorFlow / Pytorch / Chainer (DockerDesktop) ■ Optical Drive: None ■ Power Supply: [4 units] 3,000W/200V - Redundant configuration (2+2) - 80 Plus Titanium certification ■ 3-year send-back hardware warranty 16,800,000 yen (tax included)

• Desktop PC

Introducing the new rack-mounted high-performance high-end server, optimized for generative AI/LLM/deep learning development!

Our company has been manufacturing custom order BTO computers and HPC products at our domestic factory since our establishment. The Applied CERVO series is a high-end server capable of processing with top-class efficiency in the development of generative AI and LLMs (large language models) that require machine power. By adopting the highest-level CPUs and GPUs, as well as large-capacity SSDs and memory, we can create a comfortable development environment. <Lineup> ■ 2U high-end GPGPU server equipped with 2 AMD EPYC processors and 4 NVIDIA A800 GPUs ■ 4U high-end GPGPU server equipped with 2 Xeon Gold processors and 8 NVIDIA RTX 6000 Ada GPUs ■ 4U high-end GPGPU server equipped with 2 Xeon Gold processors and 8 NVIDIA A800 GPUs Additionally, we accept various customization requests based on your needs. *For more details, please refer to the PDF document or contact our sales office.

• server

A compact AI supercomputer "based on NVIDIA DGX Spark" that will revolutionize local AI development.

■ Overview The "MS-C931" is a compact AI supercomputer that condenses GPU-based high-performance computing into a small form factor. Its design, optimized for edge AI and local development environments, enables fast and secure execution of AI model training and inference without relying on the cloud. It is suitable for a wide range of applications, from research and development to prototyping and validation before product implementation. ■ Key Features - Overwhelming processing performance (equipped with GPU) It can be equipped with the latest NVIDIA GPUs, delivering computational performance of several hundred TOPS, capable of handling large-scale deep learning models. - Compact design that fits on a desk It does not require a server room and can be installed on a desk in an office or laboratory. Despite its size being comparable to a typical tower PC, it achieves high-density cooling and power design. - Ideal for AI development in local environments AI development can be conducted independently of the network, making it easy to implement in research institutions and companies with strict security requirements, eliminating cloud communication costs. It can process internal data without exposing it externally, making it particularly effective for AI development involving personal information and intellectual property.

• Industrial PCs
• Workstation
• Other production and development software and systems

From generative AI, LLM, CAE to 3DCG production. GPU solutions supporting next-generation research and development.

This lineup features NVIDIA GPU solutions that support a wide range of research and development as well as creative applications, from generative AI, LLM, CAE, to 3DCG production. It includes a broad selection from the "RTX PRO series" for workstations, "H200 NVL" for AI/HPC, the compact AI supercomputer "DGX Spark," to the high cost-performance "GeForce RTX 50 series." We propose the optimal GPU environment for research institutions, universities, manufacturing, and design development departments. ■ Example of featured products - NVIDIA RTX PRO Blackwell series - NVIDIA H200 NVL - NVIDIA DGX Spark - GeForce RTX 5090 / 5080 / 5070Ti, among others ■ Main applications - Generative AI / Local LLM development - CAE / CFD / Structural analysis - 3DCG / CAD / Video production - AI image analysis / Data science - Scientific and technical computing (HPC)

• Image analysis

HPC infrastructure to accelerate the inference and simulation of autonomous driving AI.

This is an HPC workstation developed by Applied, optimized for the NVIDIA Alpamayo development environment. It is configured to perform inference, simulation, sensor data analysis, and AI model validation required for autonomous driving system development with high precision and efficiency. Equipped with a high-performance CPU and RTX PRO 6000 Blackwell Max-Q, it supports VLA models, simulation environments, and engineering applications. It also offers installation assistance for Ubuntu 24.04 LTS and framework environment setup, providing an environment that facilitates immediate research and development after installation.

• Industrial PCs
• Desktop PC

H200-equipped server that accelerates large-scale AI/HPC with NVLink 4-Way.

This is an AI/HPC server equipped with four NVIDIA H200 NVL GPUs, supporting high-speed communication between GPUs via NVLink Bridge. While based on a standard PCIe server, it minimizes communication bottlenecks between GPUs, efficiently handling large language model inference, generative AI, real-time analytics, and scientific computing. Each GPU is equipped with a large 141GB HBM3e memory, allowing a total of 564GB of GPU memory across four GPUs. It offers Xeon 6 models and EPYC models, and can be customized to suit various applications and existing environments.

• server

A GPU environment for PhysicsNeMo that integrates physical laws and AI to accelerate design verification.

NVIDIA PhysicsNeMo is a physics AI framework that combines physical laws and AI to accelerate simulations of fluids, heat, structures, and more. This product is a GPU workstation that allows for comfortable execution of learning, inference, and what-if validation of PhysicsNeMo in a local environment. It offers a balanced model equipped with an AMD Ryzen 9 and RTX PRO 4500 Blackwell, as well as a large-scale analysis model featuring a Threadripper and RTX PRO 6000 Blackwell Max-Q. It adopts Ubuntu 24.04 LTS and supports development environments utilizing PyTorch, Docker, Google Colab, and Magnum IO. It allows for quick verification of the results of design variable changes, supporting the efficiency of design exploration and digital twin development.

• Industrial PCs
• Desktop PC

NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition equipped

APPLEID Be-Clia Workstation Type-ZU2V2-9 High-Performance Workstation NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition model

• Image Processing Equipment
• simulator
• Industrial PCs

Fukuda Co., Ltd., which develops, manufactures, and sells leak testing equipment to inspect airtightness and sealing performance, provides a technical document that clearly explains the methods and theories of airtight testing, titled "Technical Manual." We are offering it as a PDF file. Please take this opportunity to download and view it!

● What is a packaging container air leak test device? It is a leak testing device that identifies the peeling, tearing, and presence of pinholes in packaging containers for pharmaceuticals and other products, conducting a Container Closure Integrity Test (CCIT). It can be applied to a wide variety of packaging containers, including flexible packaging, bottle containers, and glass containers, primarily in pharmaceutical packaging. The measurement method employs the <sealed chamber method (pressure change leak testing method 1)>. A sample is placed in the chamber, and by applying pressure or vacuum and measuring the pressure change within the chamber, quantitative inspection is performed. It can be used in various scenarios, including offline manufacturing, inline manufacturing, CMC laboratories, quality control laboratories, and stability testing. * The containers subject to inspection are referred to as "work" or "samples." Here, specific samples are referred to as "samples" to differentiate the terminology.

The pinhole inspection for packaging engineering is fully automated, eliminating packaging defects based on certain quality standards. The judgment results and operational status can be managed via a monitor screen. - Quantitative management of quality standards for the pinhole inspection process - Ensuring the quality of the product's shelf life - Establishing evaluation criteria for the manufacturing process, achieving quantitative quality assurance - Slim and space-saving design that can be connected to existing production lines - Quick changeover for product variety switching can be handled in a short time by exchanging chambers - Fully domestically produced, allowing for rapid maintenance and troubleshooting - Example target works: pillow packaging, packaged rice (packaged cooked rice), etc. *Please consult us regarding the use of this product for pharmaceuticals.

In construction work, it is important not only to focus on the building itself but also to create a safe environment for construction vehicles and workers to move around. The Hyper CLT plywood is a lightweight wooden board that utilizes CLT. It can be used as temporary materials to support temporary roads, material delivery routes, heavy machinery work yards, material storage areas, and worker pathways, rather than as structural materials for buildings. The 5×10 board weighs approximately 230 kg, making it relatively lightweight and easy to install, remove, or relocate depending on site conditions. This allows for the boards to be moved in accordance with the progress of the construction, making it easier to effectively utilize limited space. Additionally, the natural texture of the wood blends well with construction sites in residential areas and around public facilities, contributing to considerations for metal noise and summer heat reflection. The Hyper CLT plywood is a registered technology under the Ministry of Land, Infrastructure, Transport and Tourism's NETIS system, designated as "TH-250021-A." It can be incorporated into highly public construction projects and exterior and earthworks, promoting the use of new technologies and environmental considerations. Wooden plywood, NETIS, civil engineering, CLT, lightweight wooden boards. A new option for temporary materials using wood is available for construction sites.

● The production time for casting prototypes can be shortened! Since model production is not required, the time needed to create prototypes, which previously took about a month, can be significantly reduced. ● Model costs can be reduced! By directly producing sand molds from 3D data, there is no need for model production costs. If changes to the model are necessary, they can be addressed by modifying the 3D data. ● Similar products with different shapes and dimensions can be prototyped simultaneously! Different products, such as similar shapes or variations in dimensions, can be molded at the same time in a single molding process. This allows for the verification of multiple types in a short period.