Running AI inference on Rebellions ATOM NPU with Red Hat AI
Red Hat Developer article
A deployment guide for serving large language models on Rebellions ATOM NPUs with Red Hat OpenShift AI and Red Hat AI Inference.
Red Hat
Erwan Gallen
Senior Principal Product Manager, Generative AI
Artificial intelligence infrastructure and inference
Current role
I lead Product Management for Red Hat AI Inference and hardware-accelerator enablement across the Red Hat AI portfolio.
My work covers roadmap and lifecycle management for GPUs and emerging silicon. I collaborate with engineering teams, open source communities, technology partners, and enterprise customers. Together, we bring the latest generative AI models to enterprise inference platforms. We focus on fast, cost-effective inference at scale through efficient accelerator use and Kubernetes-native distributed serving across the hybrid cloud.
Areas of focus
- AI inference
- Product management for Red Hat AI Inference, focused on fast, consistent, and cost-effective inference across the hybrid cloud.
- Hardware accelerators
- Enablement and lifecycle strategy for GPUs and emerging AI silicon.
- Hybrid cloud
- Consistent AI workloads across the Red Hat AI portfolio.
Red Hat publications and technical articles
Agentic AI demands a new infrastructure stack: AMD and Red Hat deliver
Red Hat Blog
How Red Hat AI 3.4 combines accelerator support, intelligent scheduling, and open infrastructure for enterprise agentic AI workloads.
Configure NVIDIA Blackwell GPUs for Red Hat AI workloads
Red Hat Developer article
A practical guide to enabling NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs across the Red Hat AI portfolio.
Serve and benchmark Prithvi models with vLLM on OpenShift
Red Hat Developer article
A reproducible deployment and benchmarking workflow for serving the Prithvi-EO-2.0 Earth-observation model with vLLM, Red Hat AI Inference, and Red Hat OpenShift AI.
Scaling Earth and space AI models with Red Hat AI Inference Server and Red Hat OpenShift AI
Red Hat Blog
How TerraTorch and vLLM bring Earth and space foundation models such as Prithvi-EO-2.0 into scalable production services with elastic autoscaling on Red Hat OpenShift AI.
Run cost-effective AI workloads on OpenShift with AWS Neuron Operator
Red Hat Developer article
A guide to running AI inference on AWS Inferentia and Trainium accelerators with the AWS Neuron Operator while retaining OpenShift scheduling, observability, and lifecycle management.
Introducing Red Hat AI Inference Server: High-performance, optimized LLM serving anywhere
Red Hat Blog
A technical introduction to the supported vLLM serving stack, covering parallelism, quantization, speculative decoding, model compression, accelerator choice, and portable hybrid-cloud deployment.
Accelerate model training on OpenShift AI with NVIDIA GPUDirect RDMA
Red Hat Developer article
How NVIDIA GPUDirect RDMA over Ethernet reduces communication bottlenecks and CPU copies for efficient distributed model training on Red Hat OpenShift AI.
Unlock the Power of Mistral AI with Red Hat OpenShift AI and NVIDIA DGX H100
Technical article on egallen.com
An end-to-end deployment of Red Hat OpenShift AI on NVIDIA DGX H100, covering assisted installation, storage, GPU discovery, the NVIDIA GPU Operator, accelerator-backed notebooks, model caching, and serving a Mistral large language model with Hugging Face Text Generation Inference.
Orchestrate offloaded network functions on DPUs with Red Hat OpenShift
Red Hat Developer article
An exploration of offloading OVN and OVS networking functions to NVIDIA BlueField-2 data processing units and orchestrating the resulting infrastructure services with Red Hat OpenShift.
How to install Fedora and MicroShift on the NVIDIA Jetson AGX Xavier
Technical article on egallen.com
A detailed edge-computing lab using experimental UEFI and ACPI firmware, Fedora Linux for aarch64, NVMe storage, and MicroShift on an NVIDIA Jetson AGX Xavier, followed by building an ARM64 application with Red Hat UBI and deploying it to the resulting compact OpenShift environment.
OpenShift Bare Metal provisioning with NVIDIA GPU
Technical article on egallen.com
A complete installation of Red Hat OpenShift on a seven-node Dell PowerEdge bare-metal lab, including provisioning services, networking, storage, BMC access, Node Feature Discovery, and the NVIDIA GPU Operator for Tesla T4-accelerated application workloads.
Red Hat OpenStack Platform 16.1
Technical article on egallen.com
The first part of an infrastructure series combining OpenStack, OpenShift, and OpenShift Container Storage. It prepares the RHEL-based director, network segmentation, bare-metal nodes, and overcloud deployment across a Dell PowerEdge lab running Red Hat OpenStack Platform 16.1.
Running Red Hat OpenStack Platform 16 with multiple Cells
Red Hat Blog
A practical overview of using multiple OpenStack Cells to distribute compute-resource management, reduce failure-domain size, and expand Red Hat OpenStack Platform 16 deployments.
CodeReady Containers with GPU for Data Science
Technical article on egallen.com
A compact OpenShift data-science environment built with Red Hat CodeReady Containers, PCI passthrough, Node Feature Discovery, and the NVIDIA GPU Operator, then validated with TensorFlow notebooks and comparative CPU and GPU training benchmarks.
Open Data Hub v0.5.1 release
Technical article on egallen.com
A practical evaluation of Open Data Hub 0.5.1 on Red Hat OpenShift 4.3, covering operator installation, JupyterHub and Spark customization, project creation, CPU and GPU notebook environments, and validation of accelerator access for open source data-science workflows.
NVIDIA GPU Operator with OpenShift 4.3 on Red Hat OpenStack Platform 13
Technical article on egallen.com
An OpenShift 4.3 deployment on Red Hat OpenStack Platform 13 with an NVIDIA Tesla V100 GPU worker, covering PCI passthrough, Node Feature Discovery, the GPU Operator, entitled builds, TensorFlow notebooks, nvidia-smi validation, and CPU-versus-GPU benchmarks.
OPAE with an Intel FPGA PAC Featuring Arria 10 GX
Technical article on egallen.com
A Red Hat Enterprise Linux hardware-acceleration guide for an Intel Programmable Acceleration Card with Arria 10 GX FPGA, including the OPAE SDK, kernel components, PCIe verification, bitstream programming, diagnostics, self-tests, and OpenCL workflows.
Official author profiles: Red Hat Blog and Red Hat Developer.