ML Infrastructure Engineer (ML Platform)

📋 Description

• • Architect the beating heart of CuspAI’s research engine. You will design, build and continuously evolve a cloud-native ML platform on Google Cloud Platform and Kubernetes that lets our world-leading AI chemists, physicists and materials scientists spin up massive distributed training jobs, track thousands of experiments and ship models to production—all without ever touching a YAML file.
• • Own the full MLOps lifecycle end-to-end. From source-controlled infrastructure-as-code (Terraform, Helm, Kapitan) through CI/CD, model registries, experiment tracking (MLflow, Weights & Biases or similar) and automated deployment, you will be the single source of truth for how code becomes a running, monitored, cost-optimised service.
• • Enable planet-scale distributed training. You will provision and tune multi-node GPU clusters (A100s, H100s, TPU pods) with smart checkpointing, elastic resource scaling and fault-tolerant data pipelines so that a 10-billion-parameter model can train overnight and resume gracefully if a node fails.
• • Guarantee 99.9 % uptime for the platform that powers breakthrough discoveries. Build real-time observability (Prometheus, Grafana, Alertmanager), self-healing automation and on-call playbooks so researchers sleep well while GPUs churn through exaflops of computation.
• • Optimise every dollar of cloud spend. Implement quota management, spot-instance orchestration and workload-aware bin-packing so that we can run 30 % more experiments without increasing budget—freeing cash for even bigger clusters.
• • Craft a delightful developer experience. Create opinionated SDKs, CLI tools and JupyterHub templates that abstract Kubernetes complexity, letting a chemist type `cuspai train --dataset water-filtration --gpus 64` and watch the magic happen.
• • Champion GitOps and reproducibility. Every environment—from a researcher’s laptop to production—is declared in Git, reviewed like code and rolled out automatically, ensuring that yesterday’s breakthrough can be reproduced next year.
• • Collaborate across disciplines daily. Sit shoulder-to-shoulder with ML researchers debugging convergence issues, pair with chemists optimising molecular featurisation pipelines, and sync with software engineers integrating models into customer-facing APIs.
• • Shape the strategic roadmap. As the first ML Infrastructure hire, you will define standards, pick the next tools and mentor future teammates, leaving a lasting architectural imprint on a platform that could accelerate the discovery of carbon-capture membranes, room-temperature superconductors or next-gen batteries.
• • Travel and connect. Expect quarterly trips to our London, Amsterdam or Berlin hubs to run workshops, share best practices and keep the global team aligned.