Senior AI Inference Systems Engineer

NVIDIA; Polska; On-site; PLN 293,000 - 507,000; Full time

Job summary

NVIDIA seeks highly skilled software engineers to build high-performance AI inference systems for large-scale models. You’ll architect and optimize inference stacks, GPU kernels, and compilers, while scaling workloads across multi-GPU, multi-node, and multi-cloud environments.

You’ll collaborate across inference, compiler, scheduling, and performance teams, push the frontier of accelerated computing, and contribute to benchmarks and open research.

Qualifications

• 7+ years in CS/EE with strong programming in Python and C/C++.
• Experience with performance engineering for ML frameworks.
• Familiarity with GPU programming and profiling tools.
• Experience with containers and orchestration (Docker, Kubernetes, Slurm).
• Excellent debugging, problem solving and communication.

Responsibilities

• Contribute features to vLLM and optimize inference stack performance.
• Develop and benchmark GPU kernels, build DSLs and compiler infra.
• Define benchmarking methodologies and contribute to MLPerf Inference.
• Architect scheduling/orchestration of containerized large-scale inference on GPUs.
• Publish research and integrate ideas into NVIDIA software products.

Skills

Python

C/C++

Go/Rust (plus)

Algorithms & data structures

Parallel programming

Distributed systems

Performance engineering

Communication

Education

Bachelor's in CS/CE/SE

Master's in CS/CE/SE

PhD in ML Systems / GPU arch / HPC

Tools

Docker

Kubernetes

Slurm

Nsight Systems/Compute

CUDA

MLIR/LLVM

Job description

NVIDIA seeks highly skilled software engineers to build high-performance AI inference systems for large-scale models. You’ll architect and optimize inference stacks, GPU kernels, and compilers, while scaling workloads across multi-GPU, multi-node, and multi-cloud environments.

You’ll collaborate across inference, compiler, scheduling, and performance teams, push the frontier of accelerated computing, and contribute to benchmarks and open research.