About the Role

Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems.

The AIRE Serving team is responsible for elevating the reliability of Anthropic’s token path from client to inference servers and back. The team has wide latitude to drive improvements to our expanding SaaS and product surface, uplevel reliability mindsets across Anthropic, and partner with teams internally to build more robust and reliable systems. The breadth and depth of the technical challenges someone joining this team will encounter will be career defining and we are still writing the playbooks. We are at the center of ensuring our customers have a consistently excellent experience.

Responsibilities

• Develop appropriate Service Level Objectives for large language model serving and training systems, balancing availability/latency with development velocity.


• Design and implement monitoring systems including availability, latency and other salient metrics.


• Assist in the design and implementation of high-availability language model serving infrastructure capable of handling the needs of millions of external customers and high-traffic internal workloads.


• Develop and manage automated failover and recovery systems for model serving deployments across multiple regions and cloud providers.


• Lead incident response for critical AI services, ensuring rapid recovery and systematic improvements from each incident


• Build and maintain cost optimization systems for large-scale AI infrastructure, focusing on accelerator (GPU/TPU/Trainium) utilization and efficiency

You may be a good fit if you

• Have extensive experience with distributed systems observability and monitoring at scale


• Understand the unique challenges of operating AI infrastructure, including model serving, batch inference, and training pipelines


• Have proven experience implementing and maintaining SLO/SLA frameworks for business-critical services


• Are comfortable working with both traditional metrics (latency, availability) and AI-specific metrics (model performance, training convergence)


• Have experience with chaos engineering and systematic resilience testing


• Can effectively bridge the gap between ML engineers and infrastructure teams


• Have excellent communication skills.

Strong candidates may also

• Have experience operating large-scale model training infrastructure or serving infrastructure (>1000 GPUs)


• Have experience with one or more ML hardware accelerators (GPUs, TPUs, Trainium, e.g.)


• Understand ML-specific networking optimizations like RDMA and InfiniBand.


• Have expertise in AI-specific observability tools and frameworks


• Understand ML model deployment strategies and their reliability implications


• Have contributed to open-source infrastructure or ML tooling

Logistics

Education requirements: We require at least a Bachelor's degree in a related field or equivalent experience.

Location-based hybrid policy: Currently, we expect all staff to be in one of our offices at least 25% of the time. However, some roles may require more time in our offices.

Visa sponsorship: We do sponsor visas! However, we aren't able to successfully sponsor visas for every role and every candidate. But if we make you an offer, we will make every reasonable effort to get you a visa, and we retain an immigration lawyer to help with this.