Member of Technical Staff, Runway

• 4+ years of experience running ML model inference at scale in production environments
• Strong experience with PyTorch and multi-GPU inference for large models
• Experience with Kubernetes for ML workloads—deploying, scaling, and debugging GPU-based services
• Comfortable working across multiple cloud providers and managing GPU driver compatibility
• Experience with monitoring and observability for ML systems (errors, throughput, GPU utilization)
• Self-starter who can work embedded with research teams and move fast
• Strong systems thinking and pragmatic approach to production reliability
• Humility and open mindedness; at Runway we love to learn from one another

Desirable

• We're looking for an ML infrastructure engineer to bridge the gap between research and production at Runway
• You'll work directly with our research teams to productionize cutting-edge generative models—taking checkpoints from training to staging to production, ensuring reliability at scale, and building the infrastructure that enables fast iteration
• You'll be embedded within research teams, providing platform support throughout the entire model development lifecycle. Your work will directly impact how quickly we can ship new models and features to millions of users
• Our API endpoints for real-time collaboration and media asset management is written in TypeScript, and runs in ECS containers on AWS Fargate. We leverage multiple AWS-native components, such as S3, CloudFront, Lambda, Kinesis, and SQS, as building blocks of our infrastructure
• Our inference backend is written in Python (PyTorch, TorchScript), and is deployed across multiple clusters / cloud providers. We use Kubernetes for container orchestration, and k8s-native components such as Flyte, Kueue, and Kyverno efficient job orchestration. We invest in prometheus and grafana for monitoring, and Terraform to manage our infrastructure
• Productionize model checkpoints end-to-end: from research completion to internal testing to production deployment to post-release support
• Build and optimize inference systems for large-scale generative models running on multi-GPU environments
• Design and implement model serving infrastructure specialized for diffusion models and real-time diffusion workflows
• Add monitoring and observability for new model releases—track errors, throughput, GPU utilization, and latency
• Embed with research teams to gather training data, run preprocessing scripts, and support the model development process
• Explore and integrate with GPU inference providers (Modal, E2E, Baseten, etc.)