MLOps Engineer — San Francisco, CA

About The Role

The role sits at the intersection of machine learning and infrastructure - turning experimental notebooks and one-off training scripts into reproducible, reliable, scalable pipelines that keep models shipping weekly.

This position directly shapes how fast the ML organization can iterate: the team builds the CI/CD systems, feature stores, model registries, and observability tooling that data scientists and ML engineers depend on every day.

Key Responsibilities

• Design and operate end-to-end ML pipelines - data ingestion, validation, training, evaluation, and deployment - using Kubernetes, TUxK2Dx/Dagster, and containerized workloads

• Build CI/CD workflows for ML artifacts: automated model versioning, canary deployments, and one-click rollback via MLflow or Kubeflow

• Implement monitoring and alerting for production models covering data drift, schema violations, latency, and business metric regression

• Own GPU compute cost optimization - scheduling, autoscaling, spot-instance strategies, and inference acceleration (ONNX, TensorRT, Triton)

• Partner with data scientists to productize training scripts into well-tested, parameterized pipelines with clear ownership boundaries

• Harden the ML platform for reliability and security: secrets management, access controls, disaster recovery, and multi-environment isolation

• Write runbooks, dashboards, and postmortems; participate in an on-call rotation supporting production ML systems

What We Are Looking For

• 3–6 years of experience in MLOps, DevOps, or infrastructure engineering, with at least 2 years supporting ML systems in production

• Strong Python and Bash skills; solid Kubernetes experience including Helm, ArgoCD, or equivalent GitOps tooling

• Hands-on experience with a major cloud platform (AWS, GCP, or Azure) and its ML serving/training stack

• Production experience with MLflow, Kubeflow, SageMaker, or Vertex VR for model registries and pipeline orchestration

• Track record of monitoring and debugging large-scale distributed systems; familiarity with Prometheus, Grafana, and OpenTelemetry

• Bachelor's degree in CS, Engineering, or equivalent practical experience

• Bonus: experience with Ray, Dask, or Spark for distributed training; LLM serving stacks (vLLM, Triton); feature stores (Feast, Tecton)