Senior System Engineer (Munich, Germany)

remotestar-team; Cambourne; Hybrid; GBP 90,000 - 150,000; Full time

Benefits offered by this job

Indefinite contract

Equal pay guaranteed

Variable performance bonus

Signing bonus

Relocation package

Private health insurance

Educational budget

Hybrid opportunity

Flexible working hours

High-paced cutting-edge tech

Job summary

Unknown company is seeking a senior systems engineer to design and build the control plane for bare-metal AI infrastructure. You will architect GPU scheduling across massive clusters, optimize networking for low latency, and extend Kubernetes with Operators and CRDs to expose hardware realities to data scientists.

Deep kernel and device driver work is required. You will debug complex issues from PCIe to NCCL timeouts, and define standards for drivers, firmware, and OS optimizations to maximize

Qualifications

• 10+ years of software engineering experience with Python.
• Deep Kubernetes knowledge: CRDs, Operators, API server architecture.
• Hands-on experience managing NVIDIA GPU clusters, drivers, CUDA, and NVIDIA Container Toolkit.
• Deep Linux kernel, cgroups, namespaces, and performance tuning.
• Terraform/Ansible for provisioning physical hardware.
• Experience debugging distributed systems across code, network, or silicon.

Responsibilities

• Build the Control Plane to automate bare-metal AI infrastructure lifecycles.
• Architect scheduling for large-scale GPU clusters with efficient bin-packing and gang scheduling.
• Tune the software-defined networking layer for low-latency interconnects (InfiniBand/RDMA).
• Develop Kubernetes Extensions via Operators and CRDs for hardware abstractions.
• Debug PCIe, NCCL timeouts, and kernel panics on bare-metal nodes.
• Define the Golden Image for AI workloads including drivers and firmware.

Skills

Python

Kubernetes

Linux Internals

GPU Clusters

Infrastructure as Code

Troubleshooting

Tools

NVIDIA Container Toolkit

Cluster API (CAPI)

Metal3

Tinkerbell

Canonical MaaS

OpenStack Ironic

Job description

About client :

Well-funded and fast-growing deep-tech company founded in 2019. We are the biggest Quantum Software company in the EU. They are also one of the 100 most promising companies in AI in the world (according to CB Insights, 2023) with 150+ employees and growing, fully multicultural and international.

Requirements

• Systems Programming Expertise: 10+ years of software engineering experience with strong proficiency in Python. You must be comfortable building system agents, APIs, and CLI tools.
• Deep Kubernetes Knowledge: You understand K8s internals beyond simple deployment. Experience with Custom Resource Definitions (CRDs), Operators, and the Kubernetes API server architecture.
• GPU Ecosystem Experience: Hands-on experience managing NVIDIA GPU clusters. Familiarity with NVIDIA drivers, CUDA toolkit, and the container runtime (NVIDIA Container Toolkit).
• Linux Internals: Deep understanding of the Linux kernel, cgroups, namespaces, and system performance tuning.
• Infrastructure as Code: Mastery of declarative infrastructure tools (Terraform, Ansible) but with a focus on provisioning physical hardware rather than just cloud VMs.
• Problem Solving: A proven track record of debugging complex distributed systems where the root cause could be code, network, or silicon.

Preferred qualifications

• HPC Background: Experience working with traditional supercomputing schedulers (Slurm, PBS) or modern batch schedulers (Volcano, Kueue, Ray).
• Bare Metal Provisioning: Experience with tools like Cluster API (CAPI), Metal3, Tinkerbell, Canonical MaaS, or OpenStack Ironic.
• High-Speed Networking: Knowledge of RDMA, InfiniBand, GPUDirect, and how to expose these technologies to containerized workloads.
• AI/ML Familiarity: Understanding of how distributed training works (e.g., PyTorch Distributed, Megatron-LM, DeepSpeed) and the infrastructure requirements of Large Language Models (LLMs).
• Observability: Experience building monitoring for hardware health (DCGM) and distributed tracing for long-running jobs.

Location: Applicants must have legal authorization to work in the country where the position is based

What you will be doing

• Building the Control Plane: Designing and developing the software layer (APIs, Controllers, Agents) that automates the lifecycle of bare-metal AI infrastructure.
• Orchestrating High-Scale Compute: Architecting scheduling solutions for large-scale distributed training jobs across massive clusters of GPUs (NVIDIA H200/B200/B300), ensuring efficient bin-packing and gang scheduling.
• Optimizing the Fabric: Tuning the software-defined networking layer to support low-latency interconnects (InfiniBand/RDMA/RoCEv2) essential for multi-node training.
• Developing Kubernetes Extensions: Writing custom Kubernetes Operators and CRDs to abstract complex hardware realities (topology awareness, GPU partitioning) into usable interfaces for our Data Scientists.
• Hardware-Level Debugging: Investigating and resolving deep systems issues, ranging from PCIe bus errors and NCCL communication timeouts to kernel panics on bare-metal nodes.
• Defining Standards: Creating the "Golden Image" for AI workloads, managing drivers, firmware, and OS optimizations to squeeze maximum performance out of the hardware.

Perks & Benefits

• Indefinite contract.
• Equal pay guaranteed.
• Variable performance bonus.
• Signing bonus.
• Relocation package (if applicable).
• Private health insurance.
• Eligibility for educational budget according to internal policy.
• Hybrid opportunity.
• Flexible working hours.
• Working in a high paced environment, working on cutting edge technologies.
• Career plan. Opportunity to learn and teach.
• Progressive Company. Happy people culture