Samiur Rahman is a Cloud Engineer focused on Microsoft technologies, with a background in core network engineering. He works across cloud infrastructure, AI infrastructure, and high-performance computing on Linux systems: the architecture, governance, identity, and reliability of enterprise-scale Azure environments, including production GPU inference on both cloud and on-premises NVIDIA platforms, hybrid networking, network security, and site reliability — alongside independent research in core network engineering.
He is currently a Cloud DevOps Engineer at LEO Pharma, one of the senior engineers responsible for the company’s cloud infrastructure: 200+ Azure subscriptions and a 15,000-user tenant, governed with Terraform and Azure Policy and managed through Azure CLI and Microsoft Graph. The role covers platform architecture (App Services, high-performance Linux VMs, network gateways), enterprise identity on Microsoft Entra ID — including access packages — CI/CD in Azure DevOps, Azure Monitor, hybrid networking, and reliability engineering for the production fleet.
He is the principal engineer for the company’s on-premises AI infrastructure: a seven-node NVIDIA DGX Spark (Blackwell) high-performance computing fleet in the Copenhagen data center, built from bare hardware on Ubuntu, with Azure Arc, Microsoft Entra SSO and Entra SSH. The fleet runs an on-premises Kubernetes GPU pool scheduled with Kueue, serves GPU inference with vLLM and LiteLLM, and is hardened with CIS and CrowdStrike EDR, with observability through Azure Monitor and Grafana.
He also owns the lifecycle of the company’s generative AI models in Azure AI Foundry — principally Anthropic Claude and OpenAI GPT — together with the AI gateways and model routing that connect Foundry and on-premises inference to internal systems, and architected the high-availability data gateway cluster that carries the company’s entire analytics and Power BI data flow.
He acts as the senior escalation point for incidents across the Microsoft ecosystem, leads resolution of critical network incidents, and advises Cyber Security leadership on Azure, Intune, and CrowdStrike. In 2025 he delivered around DKK 570,000 (€76,000) in annualized savings through cloud FinOps. He is part of a cross-disciplinary team of cloud, data engineering, data science, and AI, reporting to the Head of AI & Cloud Data Services in Copenhagen.
He was previously a Senior Cloud Network Engineer at Microsoft, on the Global Azure Core Network team, doing core network engineering under strict SLAs as the senior technical authority for enterprise escalations — networking problems beyond the reach of customers’ own engineering teams, across ExpressRoute, private peering, DNS, AKS, InfiniBand RDMA on H-series HPC and N-series GPU VMs, and most of the Azure networking stack, often reproducing customer failures on Red Hat Enterprise Linux. He consulted for Fortune 50 enterprises and government organizations such as the Government of the United Kingdom, the British Transport Police, and the Government of South Australia. Earlier roles were held at BNP Paribas in Kraków and Honeywell in Katowice. He holds 37 certifications from NVIDIA, Microsoft, Cisco, and Palo Alto Networks.
Outside of work he writes independent research in core network engineering, with a focus on high-performance interconnects. His 2026 working paper, From Queue Pairs to Bell Pairs, argues that InfiniBand/RDMA and quantum entanglement networks are the same engineering problem — scheduling scarce, stateful remote resources under a hard latency budget — and that the control plane of hybrid quantum–classical clusters should look like the RDMA fabrics operators already run.
Samiur is also a trainee pilot, having completed the EASA Private Pilot License ground school at ATO Avioner in Warsaw with a rating of 94%.
