Performance Engineer GPU
AI safety and research company building reliable, interpretable, and steerable AI systems, including the Claude product family and developer platform.
Funding history
About Anthropic
Anthropic PBC develops frontier AI systems and deploys them through Claude products and the Claude Platform, with a stated focus on safety, interpretability, and steerability.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will architect and implement GPU performance systems for large language models. You will optimize utilization, custom kernels, distributed communication, training and inference pipelines, memory bandwidth, and serving infrastructure. You will also build performance models, eliminate bottlenecks, and work with hardware vendors on accelerator capabilities.
Requirements
- Deep experience with GPU programming and optimization at scale
- GPU kernel development using CUDA, Triton, CUTLASS, Flash Attention, or tensor-core optimization
- PyTorch or JAX internals
- torch.compile
- XLA
- Custom operators
- Kernel fusion
- Memory-bandwidth optimization
- Nsight profiling
- NCCL
- NVLink
- Collective communication
- Model parallelism
- INT8 or FP8 quantization
- Mixed-precision techniques
- Large-scale training infrastructure
- Fault tolerance
- Cluster orchestration
- Bachelor’s degree or equivalent education, training, or experience
Responsibilities
- Maximize GPU utilization and performance at scale
- Implement GPU and systems optimizations for language models
- Develop custom kernels and mixed-precision techniques
- Design distributed communication strategies for multi-node GPU clusters
- Optimize training and inference pipelines
- Build performance-modeling frameworks for GPU utilization
- Implement kernel-fusion strategies to reduce memory-bandwidth bottlenecks
- Build resilient distributed training systems
- Profile and eliminate serving-infrastructure bottlenecks
- Partner with hardware vendors on accelerator capabilities and software stacks
Benefits
- Visa sponsorship support
- Equity donation matching
- Generous vacation
- Parental leave
- Flexible working hours