CUDA / AI Engineer
Core
Optimize LLM inference performance on modern GPUs by developing and debugging low-level CUDA kernels and analyzing GPU execution patterns.
Role type
Senior IC CUDA/AI engineer (LLM inference optimization)
Builds
High-performance LLM inference engines and serving frameworks
Domain
AI systems + High-performance computing (GPU)
Deliverable
production ML models
Required skills
CUDA programming, GPU architecture, LLM inference optimization, kernel development, performance benchmarking, memory access pattern analysis
Preferred skills
PyTorch internals, distributed inference, quantization, KV cache optimization, Triton, FlashAttention
Technologies
vLLM, SGLang, TensorRT-LLM, Triton, FlashAttention, CUDA
Responsibilities
Develop, optimize, debug, or reverse-engineer CUDA kernels; Analyze GPU execution and inference bottlenecks; Benchmark performance to improve latency and throughput; Explore optimization techniques for matrix multiplication and attention mechanisms
Seniority
Senior, hands-on IC
