Key Notes
A founding Staff SRE remit establishes reliability foundations for Wayve’s Model Development Platform and multi-tenant GPU Compute environment. Essential experience includes GPU-backed environments or large-scale ML infrastructure, production MLOps, Kubernetes, a major cloud provider, distributed systems, Linux, automation and observability. The hybrid London role requires two office days each week and participation in a 24/7 on-call rotation.
What You'll Work On
- Establish reliability, availability and performance standards for the Model Development Platform and GPU Compute services as a founding Staff SRE.
- Set SLOs, SLIs and error budgets, improving capacity planning, cluster scaling and efficiency across multi-tenant GPU fleets.
- Direct incident triage, escalation and root-cause analysis during the 24/7 rotation, converting findings into durable platform changes.
- Architect observability, production-readiness, deployment-safety and automation practices with ML, platform and software teams across large-scale cloud services.
Why This Role Matters
- Wayve’s Model Development Platform supports the route from raw data to on-road experimentation; Staff-level reliability standards shape how consistently that route operates.
- Wayve’s GPU Compute platform uses capacity planning, automation and incident recovery to keep shared training and inference workloads efficient as large compute clusters scale.
What They Are Looking For
- Experience: The Staff role requires proven SRE, Production Engineer or Cloud Reliability experience supporting large-scale cloud systems.
- Technical: Operating GPU-backed environments or large-scale ML infrastructure, plus production training or inference pipelines, is essential.
- Technical: Production Kubernetes and workloads in AWS, GCP or Azure are essential.
- Technical: Linux, proficiency in a scripting or systems language such as Python, Go or C++, distributed-systems troubleshooting and observability are essential.
- Preferred: Infrastructure as code, multi-team SLO programmes or founding an SRE function are desirable.