Atos supports NVIDIA Triton Inference Server on latest updates for AI solutions

Paris, November 9th , 2021

By orchestrating best-of-breed hardware, AI software, and expertise in digital security and consulting, Atos supports organizations and enterprises to capture data, process and run inference with the highest speed and tackling the most daunting complexity.

Atos is eager to discover the potential of the latest updates to NVIDIA Triton Inference Server, announced during NVIDIA GTC, to facilitate AI inference deployment of the Atos ThinkAI solution and Atos Computer Vision Platform for AI-based video analytics.

Atos provides a large AI inference offering from NVIDIA-Certified BullSequana X series for HPC and NVIDIA-Certified BullSequana SA20G for enterprise servers, certified for the NVIDIA AI Enterprise software suite for development, management and scaling of AI. Atos also offers the BullSequana Edge for edge inference, which is also an NVIDIA GPU-powered server.

Strategize the AI journey with the high-performance AI solution

Atos ThinkAI solution accompanies enterprises and researchers to contemplate, architect, build, and deliver scalable, energy-efficient high-performance AI solutions in weeks. The Atos ThinkAI solution includes NVIDIA DGX SuperPOD computing infrastructure, and was successfully deployed in Linkoping University in Sweden. The NVIDIA DGX A100-based supercomputer BerzeLius, named after renowned Swedish scientist Jacob Berzelius, is the most powerful supercomputer for machine learning and artificial intelligence in Europe, delivering 300 petaflops of AI performance. After its inauguration in 2021, BerzeLius is providing powerful AI computing resources to enterprises in multiple domains, such as telecommunication, pharmaceuticals, and vehicles, in collaboration with Swedish academia. In Latin America, Atos and NVIDIA delivered the Atos ThinkAI solution to Telconet for AI-empowered smart city applications.

Build, train, run and deploy AI models with Atos Computer Vision Platform

Atos Computer Vision Platform is a unique, end-to-end computer vision platform providing pre-trained and customizable AI models powered by BullSequana servers and enriched by Atos computer vision experts through worldwide experts labs. VISuite, the software toolbox, uses NVIDIA TensorRT to optimize size and speed up AI model processing without compromising on accuracy. Atos will use NVIDIA’s latest version TensorRT 8.2 for higher AI-model performance during inference and efficient solution deployment.

It enables customers to identify events and behaviors, to reduce error rates, to guarantee people and asset safety, to deliver highest quality, to offer frictionless and personalized customer experiences. Business and organizations keep up the paste of events and demand, by analyzing videos in real time at the edge to drive the best decisions.