Gpu Engineer Ai Systems Infrastructure (paris/nyc)

Company: Adamas Knight

Location: Hybrid

Posted on: April 23

We are recruiting for an impressive AI start-up backed by some of the biggest names in industry that design adaptive systems that scale with intelligence itself; enabling frontier models to learn faster, train deeper, and generalise better.

Their work spans AI research, distributed compute, and low-level systems optimisation, all built to serve models that push the limits of performance and cognition.

As a GPU Engineer, you will join a high-trust, high-impact team of engineers and researchers focused on one thing: performance. Youll work on some of the most complex GPU workloads on the planet, from kernel-level tuning to multi-cluster scaling, and build the infrastructure that powers the next generation of foundation models.

What Youll Do

• Design and optimise high-throughput GPU compute pipelines for multi-node, multi-cluster AI model training.
• Write highly optimised CUDA kernels and extend GPU runtime performance across model architectures (transformers, diffusion, etc).
• Profile and debug GPU workloads at microsecond resolution using low-level tools (e.g., Nsight Compute, NVTX, CUPTI).
• Develop and maintain custom collective communication primitives with NCCL or equivalent for high-bandwidth, low-latency scaling.
• Co-engineer with hardware teams to tune GPU utilisation across H100, custom silicon, or bleeding-edge accelerators.
• Build tools and monitoring infrastructure to surface latency, memory fragmentation, kernel bottlenecks, and PCIe/NVLink performance.
• Push compiler/runtime boundaries (e.g., Triton, MLIR, XLA) to get closer to bare-metal performance.
• Contribute to model-parallel, tensor-parallel, and pipeline-parallel systems for efficient LLM training.

What You'll Need To Bring

• 2+ years of hands-on experience with CUDA and GPU performance engineering in production environments.
• Deep knowledge of GPU architecture (H100/A100 or comparable), including memory hierarchies, streaming multiprocessors, and interconnects.
• Proficiency with PyTorch or JAX and hands-on experience with large-scale training or distributed model serving.
• Strong C++ and Python skills, with a systems mindset and passion for reliable, testable code.
• Experience profiling, debugging, and optimising under real-world training loads.
• Exposure to cluster orchestration, job schedulers, or GPU-aware resource managers is a bonus.
• Comfortable navigating both startup-level pace and production-grade systems discipline.

What They Offer

• Work on problems that matter, at world-changing scale;
• Competitive compensation and equity packages;
• Hybrid/onsite work flexibility in Paris or NYC;
• Visa sponsorship and relocation assistance for international candidates;
• Comprehensive medical insurance;
• 401(k) plan with 4% matching (or equivalent);
• Unlimited PTO, strongly encouraging 5 or more weeks;
• Dedicated budget for mental wellness, personal growth, and overall well-being;
• & much more...

At Adamas Knight, we are committed to creating an inclusive culture. We do not discriminate based on race, religion, gender, national origin, sexual orientation, age, veteran status, disability, or any other legally protected status. Diversity is highly valued, and we encourage applicants from all backgrounds to apply.