Software Engineer, Early Careers AI/UI

Roku•San Jose, CA

2d•Hybrid

About The Position

Roku is seeking engineers for its OS Platform UI team to build real agentic capability into the platform behind over 100 million streaming households. This is a systems engineering job focused on integrating agent work with telemetry, distributed services, and physical devices. The role requires engineers who understand the complexities of shipping schema changes to millions of devices and can make agents work on systems that already handle significant load. The Roku OS Platform UI team is responsible for the user experiences that greet every Roku user, including the home screen, Roku Channel, and other content discovery features. Engineers on this team own features end-to-end, from the OS platform layer through application and UI to cloud integration, collaborating with product, design, and cross-functional engineering partners to deliver ambitious ideas at scale.

Requirements

Up to 3/4 years of professional engineering experience since completing studies.
Strong systems or infrastructure fundamentals, including real-time data, telemetry, distributed systems, platforms, or device work.
Experience building and shipping something agentic that other people use (work or personal project), involving multiple steps beyond a simple chat window wrapper.
Ability to clearly articulate how you stay accountable for code you did not type, demonstrating a design-first approach, review discipline, and personal gates.
A perspective on the current limitations of agents and areas for improvement.
Existing right to work in the U.S. (visa sponsorship is not available for this role).

Nice To Haves

Experience working fluently with AI tools.

Responsibilities

Develop multi-step agents that perform real work in production, including detecting regressions, bisecting commits, proposing fixes, and generating code for review.
Build internal tooling, skills, harnesses, and services that enhance team efficiency and adoption.
Process real-time telemetry and events at streaming scale, adhering to CPU and memory constraints of consumer devices.
Address agent problems that arise specifically with hardware in the loop, such as physical feedback, device fleets, and signals not reproducible on a laptop.