Senior Software Engineer, GPU Infrastructure (HPC)

About the role

The internal infrastructure team is responsible for building world-class infrastructure and tools used to train, evaluate and serve Cohere's foundational models. By joining our team, you will work in close collaboration with AI researchers to support their AI workload needs on the cutting edge, with a strong focus on stability, scalability, and observability. You will be responsible for building and operating superclusters across multiple clouds. Your work will directly accelerate the development of industry-leading AI models that power Cohere's platform North.

Note: All of our infrastructure roles require participating in a 24x7 on-call rotation, where you are compensated for your on-call schedule.

Responsibilities

• Build and scale ML-optimized HPC infrastructure: Deploy and manage Kubernetes-based GPU/TPU superclusters across multiple clouds, ensuring high throughput and low-latency performance for AI workloads.
• Optimize for AI/ML training: Collaborate with cloud providers to fine-tune infrastructure for cost efficiency, reliability, and performance, leveraging technologies like RDMA, NCCL, and high-speed interconnects.
• Troubleshoot and resolve complex issues: Proactively identify and resolve infrastructure bottlenecks, performance degradation, and system failures to ensure minimal disruption to AI/ML workflows.
• Enable researchers with self-service tools: Design intuitive interfaces and workflows that allow researchers to monitor, debug, and optimize their training jobs independently.
• Drive innovation in ML infrastructure: Work closely with AI researchers to understand emerging needs (e.g., JAX, PyTorch, distributed training) and translate them into robust, scalable infrastructure solutions.
• Champion best practices: Advocate for observability, automation, and infrastructure-as-code (IaC) across the organization, ensuring systems are maintainable and resilient.
• Mentorship and collaboration: Share expertise through code reviews, documentation, and cross-team collaboration, fostering a culture of knowledge transfer and engineering excellence.

Qualifications

• Deep expertise in ML/HPC infrastructure: Experience with GPU/TPU clusters, distributed training frameworks (JAX, PyTorch, TensorFlow), and high-performance computing (HPC) environments.
• Kubernetes at scale: Proven ability to deploy, manage, and troubleshoot cloud-native Kubernetes clusters for AI workloads.
• Strong programming skills: Proficiency in Python (for ML tooling) and Go (for systems engineering), with a preference for open-source contributions over reinventing solutions.
• Low-level systems knowledge: Familiarity with Linux internals, RDMA networking, and performance optimization for ML workloads.
• Research collaboration experience: A track record of working closely with AI researchers or ML engineers to solve infrastructure challenges.
• Self-directed problem-solving: The ability to identify bottlenecks, propose solutions, and drive impact in a fast-paced environment.

Benefits

• A weekly lunch stipend of $75/£75 or equivalent in your local currency for lunch.
• Full health and dental benefits, including a separate budget for mental health.
• RRSP matching, 401K, Pension Scheme.
• 100% Parental Leave top-up for up to 6 months, for either parent.
• Annual enrichment benefits: Arts & culture, fitness/wellness, quality time, and a workspace improvement credit.
• Education & learning stipend for conferences, courses, and coaching.
• 6 weeks of paid vacation (30 working days!)
• Budget for traveling to other offices if you are remote, plus an annual company offsite.

Pay

CA$285K - CA$340K

Schedule

Remote-friendly with offices in Toronto, London, New York City, San Francisco, Montreal, Paris, Berlin and Seoul. For those in the office: a daily lunch program, plenty of snacks, and regular community and social events. For those not near an office: a co-working benefit so you can work alongside others in your city. Everyone receives a $500 home office stipend to set up your workspace properly.