Software Engineer, Cloud + AI

About the role

Software Engineer, Cloud + AI

Swish Robotics · San Francisco · On-site · Full-time

Swish builds a countertop robot that actually cooks. You put in a pod or your own ingredients and it hands you a finished meal. The first units are in customers' kitchens now, and every one of them talks to AWS every day.

We are very early with a functional robot, real users, and a lot of the system still to be built. This role is for an engineer who wants to shape the cloud and AI side of the product while it is still small enough to shape.

Where you would contribute

The fleet platform on AWS. The cloud that every robot depends on: how units are identified and updated in the field, how we keep development and production apart, how we know something is wrong before a customer does, and how we see what the fleet is doing.

The data platform. Every cook produces data. Some of it lands in S3 today; there is a lot more to do to make it queryable and to turn it into training and evaluation sets.

AI in the loop. We want the robot to cook well from what it senses rather than from a hand-tuned model per dish. That needs training pipelines, evaluation, packaging for the device, delivery over the air, and a feedback loop from the fleet back into training.

Product features that live in the cloud. The app, the customer experience, and the tools our culinary team uses all have a backend. Small features, real users, fast turnaround.

Units in homes. When a customer's robot misbehaves, someone looks at its logs and decides what to ship. Everyone here is on deployment duty.

What we need from you

You have shipped and operated production systems on AWS, and you have done it with infrastructure as code (CDK, Terraform, or CloudFormation). You have worked with connected devices, or something close enough that IoT Core, MQTT, device shadows and OTA are not new words to you.

You write Python well. You are comfortable on a Linux box you cannot see: systemd, journals, Ansible, a device that reboots mid-command.

You take security seriously without being precious about it: PKI, secrets management, least privilege, what a stolen device can and cannot leak.

You can take a trained model and make it real: data pipelines, evaluation that means something, packaging for an edge device, a rollout you can roll back. You do not need to be a researcher. You need to be the person who makes the researcher's work run on every machine.

You have worked somewhere small enough that you owned outcomes, not tickets.

Nice to have

Flutter or Dart. Embedded experience. Computer vision or sensor fusion. Observability and analytics tooling.