Expression of Interest (Member of Technical Staff)

Berkeley, CA /
Technical /
/ On-site
[Not currently hiring, though applicants will be evaluated in the future.]

We are working on various engineering-heavy ML-based alignment projects, like this paper we recently published on High-stakes Alignment via Adversarial Training. We’re hiring for technical staff to contribute to these projects, doing some combination of machine learning, engineering and reasoning about AI alignment.

We are accepting applications for full time positions and internships.

On the ML research side, responsibilities include:

    • Designing, running, and analyzing experiments
    • Proposing and testing hypotheses for improving model performance
    • Reasoning about how to apply ML algorithms to new situations, and about what techniques might be expected to scale well with future capabilities

On the engineering side, responsibilities include:

    • Building infrastructure required to make our research productive, e.g. infrastructure for running large hyperparameter searches on the cloud. 
    • Solving various software problems quickly, eg building a web app for data collection and quickly refining it based on human feedback.
    • Building out our codebases for ML projects in ways that maintain the robustness of the code while improving the flexibility of the experiments that we can run with it.

We're looking for candidates with the following properties:

    • Are comfortable switching between responsibilities rapidly as the needs of the project evolve.
    • Have enthusiasm for thinking strategically about how we should execute on our plans. (Everyone on our technical staff has at some point had primary responsibility for project management of the main project, just because they happened to have the most context on the core problems at the time.)
    • Have broad knowledge of many topics in computer science, math, and machine learning, and have enthusiasm for quickly picking up new topics.
    • Are dedicated to doing the alignment research that will most increase the expected value of the long-term future.
We welcome applications from people who don’t fulfill all of these requirements. This is a full-time role and we expect staff to work in-person in our office in downtown Berkeley. We will consider international applicants who need visas. Salaries will vary based on experience and be roughly competitive with FAANG tech salaries.

Redwood Research is led by Nate Thomas (CEO), Buck Shlegeris (CTO), and Bill Zito (COO), and a board of Nate Thomas, Holden Karnofsky, and Paul Christiano.