Research Engineer

Imbue•San Francisco, CA

About The Position

Imbue's mission is to make "open" AI agents win over closed agents. To do that, we’re starting with making open coding agents, since they both make it easier to construct future agents, and we can use them ourselves. Our flagship product, Sculptor, gives engineers the power to run and coordinate multiple coding agents in parallel, helping them move faster and stay in flow. At its core, Sculptor is about giving developers power to make software creation easier, faster, and more reliable. We intend to make the source code available to both serve as an example of how to create open agents, and provide a tool that makes it easier for others to create open software. The insights we gain from building Sculptor help us both understand the interaction design and infrastructure necessary for creating open agents, and improve the capabilities of the underlying models. Ultimately, we aim to rekindle the dream of the personal computer, where computers become truly intelligent tools that empower us, giving us freedom, dignity, and agency to pursue the things we love. In this role, you’ll build the AI systems that power Sculptor and ensure that coding agents work “for you”. You’ll directly shape the intelligence behind the product you use every day, and will have real influence on the problems we tackle and the approaches we take. We're looking for scientists who combine research excellence with pragmatic engineering; someone who can push the boundaries of what's possible while shipping real solutions that delight our users.

Requirements

Driven by making AI agents that actually work in the real world
Excited to see your research ship to thousands of developers
Independent and self-motivated
Comfortable with Python, LLMs, and agentic approaches

Responsibilities

Build verification systems that make coding agents more reliable, including test generation and LLM-as-judge approaches
Turn ambiguous problems into quantifiable ones, through a mixture of annotation, synthetic data generation, and product logging
Create agent improvement loops via self critique and evolutionary algorithms
Build experiences and tooling that help developers easily audit and manage coding agents
Partner with the product team to translate cutting-edge applied research into features users actually want