Palk täpsustamataTuruvahemik: 4250–5800 €
Responsibilities
Optimizing LLM inference performance on modern GPUs, including analyzing GPU execution, memory access, and kernel bottlenecks; developing, debugging, or reverse-engineering CUDA kernels and low-level GPU operations; and benchmarking improvements in latency, throughput, and GPU utilization.
Requirements
- CUDA
- GPU programming
- GPU architecture
- LLM inference
- vLLM
- SGLang
- TensorRT-LLM
- Triton
- FlashAttention
- PyTorch internals
Employer offers
- Flexible working hours
- 29 days PTO
- 10 paid recovery days
- Legal support
- Financial support
- English classes
- HR support
- Professional development