1 open role mentioning Accelerator programming
Anthropic
Anthropic's inference engine is the core software responsible for managing the entire token path, from request batching and model distribution across chips to memory management and coordinating forward passes. This in-house system serves Claude to millions of users and supports research workloads across all accelerator platforms. The role involves building and optimizing this system at scale, focusing on improving throughput, cost, reliability, and latency. It requires a deep understanding of hardware and bandwidth metrics, with the ability to quickly model performance bottlenecks. This is a highly technical and impactful position for engineers who enjoy working with accelerator programming, high-performance host-device systems, and large-scale distributed systems. Familiarity with transformer architecture is beneficial.
$350k - $850k
