GPU Engineer (AI)

TL;DR

GPU Engineer (AI): Building the fastest LLM inference engine on standard datacenter GPUs with an accent on low-level GPU optimization and monokernel development. Focus on writing optimized CUDA/HIP kernels, implementing inter-GPU collectives, and scaling the stack for MoE models.

Location: Hybrid (Paris, France) - Must spend at least 50% of time in the office

Company

hirify.global builds the fastest LLM inference engine on standard datacenter GPUs.

What you will do

• Contribute to a monokernel pipeline covering the full decode pass across AMD and NVIDIA architectures.
• Develop low-level GPU optimizations, including grid synchronizations and inter-GPU collectives.
• Create optimized GEMM and attention kernels for specific batch sizes and context lengths.
• Build profiling infrastructure with custom instrumentation and device-timestamp frameworks.
• Scale the stack to support third-party MoE models such as DeepSeek v4 and Qwen 3.
• Develop AI agents for autonomous GPU engineering research and kernel optimization.

Requirements

• Proven experience writing GPU kernels where performance was the central constraint (code samples required).
• Deep understanding of hardware below the framework level, including inline PTX or CDNA ISA.
• Knowledge of latency-sensitive execution paths and GPU internals.
• Must be based in Paris, France, or able to be present in the office at least 50% of the time.
• A degree from a top engineering school or a PhD with concrete GPU work.

Culture & Benefits

• Direct access to AMD and NVIDIA datacenter GPUs from day one.
• High-impact environment where technical judgment and creativity shape key decisions.
• Compensation aligned with top technical profiles in the Paris AI market, including equity.
• Opportunity to solve critical problems directly influencing model execution speed.