ML Systems Engineer for High-Performance AI Inference

Arago; Paris; On-site; EUR 90,000 - 140,000; Full time

Benefits offered by this job

Stock options

Healthcare

Pension

Professional development

25 days PTO

Job summary

Arago in Paris is seeking engineers to optimize the execution of modern AI models on Arago’s custom accelerator. You will work across kernels, model execution, multi-device distribution, runtime, and inference serving, shaping the software stack around the accelerator's capabilities.

You will leverage CUDA, Triton, ROCm/HIP, and related tools to accelerate workloads, with strong C++ and Python skills. English proficiency is required for collaboration across teams.

Qualifications

• Strong experience in high-performance ML inference or ML systems engineering.
• Deep understanding of computer architecture and memory hierarchies.
• Experience developing/custom kernels with CUDA, Triton, ROCm/HIP or equivalents.
• Experience with operator fusion, tiling, scheduling, data movement optimization, graph execution.
• Strong understanding of distributed model execution and parallelism.
• Hands-on with modern inference-serving systems (e.g., vLLM, KV-cache management).
• Strong C++ and Python skills and exposure to MLIR is a plus.
• Language: English at a proficient level.

Responsibilities

• Analyze kernels, runtime, memory, and system bottlenecks on Arago's accelerator.
• Develop and optimize custom kernels and execution strategies.
• Map models and operators across multiple Arago devices with efficient communication.
• Develop inference-serving techniques like continuous batching and KV caches.
• Build profiling and performance-analysis infrastructure for kernels and models.
• Collaborate with hardware, compiler, and runtime teams to influence future features.

Skills

ML inference

GPU programming

Distributed execution

C++

Python

Inference serving

Profiling tools

Tools

CUDA

Triton

ROCm/HIP

TensorRT-LLM

MLIR

Job description

Arago in Paris is seeking engineers to optimize the execution of modern AI models on Arago’s custom accelerator. You will work across kernels, model execution, multi-device distribution, runtime, and inference serving, shaping the software stack around the accelerator's capabilities.

You will leverage CUDA, Triton, ROCm/HIP, and related tools to accelerate workloads, with strong C++ and Python skills. English proficiency is required for collaboration across teams.