Anthropic is seeking a Linux OS and System Programming Subject Matter Expert to join our Infrastructure team. In this role, you'll work on accelerating and optimizing our virtualization and VM workloads that power our AI infrastructure. Your expertise in low-level system programming, kernel optimization, and virtualization technologies will be crucial in ensuring Anthropic can scale our compute infrastructure efficiently and reliably for training and serving frontier AI models.
Responsibilities:
- Optimize our virtualization stack, improving performance, reliability, and efficiency of our VM environments
- Design and implement kernel modules, drivers, and system-level components to enhance our compute infrastructure
- Investigate and resolve performance bottlenecks in virtualized environments
- Collaborate with cloud engineering teams to optimize interactions between our workloads and underlying hardware
- Develop tooling for monitoring and improving virtualization performance
- Work with our ML engineers to understand their computational needs and optimize our systems accordingly
- Contribute to the design and implementation of our next-generation compute infrastructure
- Share knowledge with team members on low-level systems programming and Linux kernel internals
- Partner with cloud providers to influence hardware and platform features for AI workloads
You may be a good fit if you:
- Have experience with Linux kernel development, system programming, or related low-level software engineering
- Understand virtualization technologies (KVM, Xen, QEMU, etc.) and their performance characteristics
- Have experience optimizing system performance for compute-intensive workloads
- Are familiar with modern CPU architectures and memory systems
- Have strong C/C++ programming skills and ideally experience with systems languages like Rust
- Understand Linux resource management, scheduling, and memory management
- Have experience profiling and debugging system-level performance issues
- Are comfortable diving into unfamiliar codebases and technical domains
- Are results-oriented, with a bias towards practical solutions and measurable impact
- Care about the societal impacts of AI and are passionate about building safe, reliable systems
Strong candidates may also have experience with:
- GPU virtualization and acceleration technologies
- Cloud infrastructure at scale (AWS, GCP)
- Container technologies and their underlying implementation (Docker, containerd, runc, OCI)
- eBPF programming and kernel tracing tools
- OS-level security hardening and isolation techniques
- Developing custom scheduling algorithms for specialized workloads
- Performance optimization for ML/AI specific workloads
- Network stack optimization and high-performance networking
- Experience with TPUs, custom ASICs, or other ML accelerators
Representative projects:
- Optimizing kernel parameters and VM configurations to reduce inference latency for large language models
- Implementing custom memory management schemes for large-scale distributed training
- Developing specialized I/O schedulers to prioritize ML workloads
- Creating lightweight virtualization solutions tailored for AI inference
- Building monitoring and instrumentation tools to identify system-level bottlenecks
- Enhancing communication between VMs for distributed training workloads
Deadline to apply: None. Applications will be reviewed on a rolling basis.
