About the role:
The mission of the Capacity Engineering & Efficiency team is to provide input into our company-wide cloud infrastructure strategy and efficiency deliverables, with a specialized focus on ML Scheduling and Observability for our Compute infrastructure. You will develop and optimize scheduling systems for our large-scale machine learning workloads, particularly working with our Python-based scheduling architecture and orchestrating workloads across jobs. Your work will contribute to our path toward building RL-aware schedulers while supporting and improving our model development through improved observability and capacity efficiency. You will be expected to work with engineering teams to ensure optimal operation and growth of our infrastructure from both a cost and technology perspective, collaborate with research engineering to scope and understand the observability and capacity needs for model development, and partner cross-functionally with finance and data science teams to analyze and forecast growth.
You may be a good fit if you:
- Experience instrumenting ML workloads for performance monitoring/efficiency
- Experience with high performance, large scaled distributed systems
- Experience with LLM inference and Reinforcement Learning
- Observability tooling and best practices (logging, metrics, tracing)
- 5+ years experience in capacity efficiency or performance engineering
- 5+ years experience in a technical role
- Have experience in scripting and building automation tools
- Are self-disciplined and thrives in fast paced environments
- Have Excellent communication skills
- Pick up slack, even if it goes outside your job description
- Have attention to detail and a passion for correctness
Strong candidates may also have experience with:
- Reinforcement Learning
- Cross-Platform accelerators
- Pytorch
- Python
- Kubernetes
- Performance optimization across multiple platforms/environments
Representative projects:
- Develop self-service tools and dashboards to enable anthropic engineers to understand their capacity, efficiency, and costs, leveraging observability best practices
- Investigate capacity requests and recommend right-sizing strategies for performance optimization across multiple platforms/environments
- Design and implement observability solutions that provide insights into infrastructure efficiency for large-scale distributed systems
- Collaborate with engineering teams to identify and resolve performance bottlenecks in Kubernetes-based ML infrastructure
- Partner with research teams to quantify computational requirements for new ML initiatives and develop appropriate capacity plans
Deadline to apply: None. Applications will be reviewed on a rolling basis.
