Software Engineer, Infrastructure: GCP, Kubernetes & GPU

Descript's Platform team owns the base the company runs on: compute and deployment, reliability and on-call, CI/CD, developer environments, the GPU infrastructure for model training and inference, and its security. Your users are Descript's other engineers.

Key responsibilities

• Run the platform: GCP, Kubernetes, Temporal, the GPU fleet and the deploy and rollback tooling
• Own GPU capacity, training and inference pipelines, and their reliability and cost
• Treat cost as an engineering limit and make clear trade-offs
• Own security for the systems you run: access, secrets, least privilege and supply chain
• Write self-explaining infrastructure as code, runbooks and observability, and join the on-call rotation

Requirements

• 8+ years building and running production distributed systems
• A major cloud and Kubernetes in production, with infrastructure as code by default
• Incident command experience, and SLOs and error budgets used in practice
• Owned an architecture or migration from plan to launch
• Uses AI agents to multiply impact

Nice to have

• GPU and ML infrastructure: capacity planning, training or inference pipelines
• Security engineering: IAM, secrets, supply chain
• Cloud cost modelling and CI/CD at monorepo scale

Remote details

• Remote role in the United States; Descript hires for remote and hybrid roles
• Remote staff meet in person a few times a year

Salary and benefits

Base salary $220,000 to $292,000 a year, plus equity. Benefits include healthcare, 401(k) matching and flexible vacation.

How to apply

Descript promises a human review for any application that includes a share link to a video made in Descript. Apply on the company's own site.