What you'll be doing

Designing and building self-service platforms for model training, distributed training, and inference on shared GPU infrastructure
Developing APIs, SDKs, and workflows that automate workload submission, infrastructure provisioning, and execution
Building orchestration and distributed execution capabilities using Flyte, Ray, or equivalent technologies
Implementing workload scheduling, queueing, resource allocation, and autoscaling across Kubernetes-based GPU environments
Building reliable workload lifecycle management, including checkpointing, retries, failure recovery, and cancellation
Establishing monitoring, multi-tenant isolation, and cost visibility to improve platform reliability and GPU utilisation

Requirements

Impressive Python engineering skills, and a track record of designing, building, and operating scalable backend platforms in production
Experience building or extending ML platforms for training or inference on GPU infrastructure
Expertise with Kubernetes, containerised workloads, and cloud infrastructure
Familiarity with workflow orchestration and distributed execution frameworks, such as Flyte, Ray, or equivalent technologies
An understanding of distributed training, GPU resource allocation, workload scheduling, and autoscaling
Experience designing reliable distributed systems with fault tolerance, observability, and operational ownership
A degree in a STEM subject or equivalent practical experience, with solid computer science fundamentals and software engineering practices