Applied AI Engineer

Role Overview

You'd work on the questions underneath an AI product: what should the system measure, what evidence should it trust, and how do we know when a change made it better? The work may involve model behavior, retrieval, agents, human feedback, or evaluation design, depending on the team and the problem in front of you.

This is applied research. You would read enough to understand the state of the art, design experiments that answer a real question, and build the evaluation or product machinery needed to learn from the result. That may mean agentic observability, Temporal workflows, evaluation harnesses, traces, model behavior, or the gap between a benchmark and what happens in actual use. A good researcher here is as comfortable explaining why an idea failed as presenting the idea that worked.

Who thrives here:

• You have strong foundations in machine learning, statistics, computer science, or a neighboring field, and you can reason clearly about evidence.

• You use Python and modern model tooling comfortably, but you are more interested in the question than in attaching the newest framework to it.

• You know that a compelling demo is not the same thing as a reliable system. You think about evaluations, failure modes, and calibration early.

• You can turn a vague research question into a small experiment, a useful baseline, and a decision about what to try next.

• You want your work to reach production and change how people use a product, not stay in a notebook because the result is interesting.

• You enjoy the strange part of the work, where the system behaves differently at 2am, the trace tells half the story, and the right answer is still being worked out.

• You can move between a paper, a terminal, and a conversation with an engineer or product person without losing the thread.

• You follow the research and builder community on Twitter/X, read beyond the paper abstracts, and have a point of view about where the field is heading.

Some nice-to-haves:

• Evaluation of open-ended or agentic systems

• Experience with retrieval, fine-tuning, or inference systems

• Work on safety, fairness, interpretability, or reliability

• A paper, thesis, open-source project, or experiment you can explain clearly

• Public work you can point to, whether that is writing, talks, research notes, or a thoughtful online presence

• Technical writing, research notes, or an active presence in the AI community online

Why this role?

AI teams are still deciding which problems deserve research and which ones need better engineering. That leaves room to shape the questions, the methods, and the standard for what counts as working. You get to see the loop from an idea to an experiment to a product decision while working close to the frontier of machine learning in San Francisco.

This is a seat for someone who wants meaningful equity, real influence over the research agenda, and unfettered experimentation with models, agents, evals, and the systems around them. You will have access to serious compute and technical tools, health coverage, flexible time off, and relocation support when the team is building in San Francisco. The work can shape products an entire generation will use, and you will help decide what those products should be able to do.