Product Engineer (Backend)

What you'll work on

• The voice pipeline. Scale real-time transcription so notes land accurately in seconds, on every device people use.

• New models in production. Evaluate new voice and transcription models as they're released, and get the best ones into users' hands.

• The agent harness. Build the harness Granola's agents run in: long-running jobs, tool calls, retries, and retrieving the right context.

• APIs and integrations. Extend the APIs and integrations that let people bring Granola context into the AI tools they already use.

• Enterprise-grade trust. Build the permissions, data isolation and admin controls that let a company roll Granola out across every team.

• Data at scale. Own Postgres performance, observability and alerting as usage grows.

Who we're looking for

We don't set a minimum number of years. Some of our strongest engineers graduated two years ago; others have run their own companies. We care about what you've built and how you think about it.

• You've owned backend systems in production that real users depended on, and you can talk through the trade-offs you made.

• You're strong in TypeScript and Node.js.

• You know relational databases deeply (we use Postgres): query plans, indexing, and migrations without downtime.

• You're comfortable with cloud infrastructure as code (we use AWS, Terraform and a lot of Lambda).

• You think about security and failure modes before the happy path.

• You have product sense: you can explain why a system matters to the person using it, and you'd rather ship what users need this week than the perfect version next quarter.

What we offer

We're living in the most exciting time for tool builders since Engelbart's demo in 1968, and we want to build that future together in London. We pay slightly above market on salary and above market on equity. We work together in person five days a week at our office in Old Street, and we offer relocation support if you're moving to London.

Great people come from all kinds of backgrounds. If this speaks to you, whether or not you look like a fit on paper, we'd love to hear from you.