Research Engineer - CUDA Kernel Engineering

Develops and optimizes CUDA kernels for AI models accelerating semiconductor design, verification, and chip optimization across large-scale GPU clusters. Integrates with frameworks like PyTorch and latest NVIDIA hardware for training, inference, and RL workloads.

About the job

Responsibilities

Develop, integrate, and optimize state-of-the-art CUDA kernels to power AI models accelerating semiconductor design and verification.
Enable large-scale model training, inference, and reinforcement learning systems for circuit layouts, RTL generation/validation, and chip architecture optimization across thousands of GPUs.
Build tools, performance benchmarks, and integration layers to push GPU utilization limits for compute-intensive workloads in AI-driven hardware design.
Collaborate with researchers and engineers; contribute kernels and tooling to open-source AI and HPC ecosystems.

Requirements

Writing and optimizing CUDA kernels for large-scale AI workloads (attention, routing, graph-based operations, physics-inspired operators).
Profiling and optimizing GPU performance for custom compute or memory-bound workloads.
Integrating custom kernels into training and inference frameworks (PyTorch, Megatron, vLLM, TorchTitan).
Working with latest NVIDIA hardware/software (Hopper, Blackwell, NVLink, NCCL, Triton).
Building GPU-accelerated primitives for graph reasoning, symbolic computation, or hardware simulation.
Collaborating with AI researchers and semiconductor experts to translate domain-specific workloads into high-performance GPU code.