11 months ago
Base Salary
$315k - $560k/yr
Responsibilities
- Architect and implement foundational GPU performance and systems infrastructure for Claude and large language models.
- Develop custom kernels, tensor-core optimizations, kernel fusion strategies, and quantization techniques.
- Optimize end-to-end training, inference, and production serving pipelines.
- Design distributed communication strategies and orchestrate large multi-node GPU clusters.
- Build performance modeling frameworks and profile and eliminate production bottlenecks.
- Create resilient systems for large-scale distributed training and partner with hardware vendors on future accelerator capabilities and software stacks.
Requirements
- Deep experience with GPU programming and optimization at scale.
- A track record of delivering transformative GPU performance improvements in production ML systems.
- Ability to navigate systems from hardware interfaces through high-level ML frameworks.
- Experience with GPU kernel development, ML compilers and frameworks, performance engineering, distributed systems, low-precision techniques, or production-scale training infrastructure is valued.
- Bachelor's degree in a related field or equivalent experience.
- Strong collaboration, communication, problem-solving, and ability to work effectively in ambiguous environments.
Benefits
- Hybrid policy requiring staff to work from an office at least 25% of the time, with some roles requiring more office time.
- Visa sponsorship may be available, with immigration-lawyer support.
- Competitive compensation, optional equity donation matching, generous vacation and parental leave, flexible working hours, and office collaboration space.
About Anthropic
We're an AI research company that builds reliable, interpretable, and steerable AI systems. Our first product is Claude, an AI assistant for tasks at any scale. Our research interests span multiple areas including natural language, human feedback, scaling laws, reinforcement learning, code generation, and interpretability.