Research Scientist, Frontier Red Team

About the role

What you’ll do

• Design capability evaluations that measure real uplift, with threat models, rubrics and baselines grounded in domain expertise.
• Elicit the upper bound of model ability using scaffolding, tool access, fine-tuning and best-of-n sampling, and report where the estimate stays uncertain.
• Build automated evaluation harnesses, including long-horizon agentic tasks in sandboxed environments.
• Work with external domain experts under appropriate controls to validate findings and calibrate severity.
• Write safety-case evidence that states clearly what was tested, what was not, and how confident we are.
• Brief leadership and release reviewers on results, including uncomfortable ones.

What you’ll bring

• Research experience in machine learning, security, or a relevant scientific domain, with a record of careful empirical work.
• Experience designing experiments and evaluations, including awareness of contamination, sandbagging and elicitation gaps.
• Strong Python skills and comfort building agent scaffolds and evaluation infrastructure.
• Depth in at least one risk area, such as vulnerability research, applied biosecurity or autonomous systems.
• Statistical discipline: confidence intervals, effect sizes and honest reporting of null results.
• Sound judgement with sensitive information and a willingness to hold results back when that is the right call.

Nice to have

• Experience in offensive security competitions or professional penetration testing.
• A graduate degree or hands-on laboratory background in a life science.
• Prior work on model organisms of misalignment or deception evaluations.
• Publications on evaluation methodology or dangerous-capability measurement.

Work you might take on

1. 01

   Build a long-horizon cyber evaluation where an agent must find and exploit vulnerabilities in a realistic, isolated network.

2. 02

   Measure how much scaffolding and repeated sampling raise scores on a hazardous-knowledge benchmark compared with a single zero-shot attempt.

3. 03

   Design a test for whether a model can acquire compute, money or access without human help in a sandboxed setting.

4. 04

   Develop a rubric and expert-review process that turns open-ended model outputs into calibrated uplift estimates.

Compensation

$300K – $500K

The US base salary range for this role. Every offer also includes equity in Clexerl and, when it spins out, in the venture you help build. We tell you the reasoning behind your number.

Benefits

• Health, fully covered

  Medical, dental and vision for you and your dependants, premiums paid in full.

• Two kinds of equity

  Equity in Clexerl, plus a stake in the venture you help build when it spins out.

• 22 weeks parental leave

  Fully paid, for every parent, however your family arrives.

• Rest that sticks

  Flexible time off with a four-week minimum we actually check on, plus a company-wide week in December.

• Learning and compute

  A yearly budget for courses, conferences and books, and compute credits for side projects.

• 401(k) with a 4% match

  Matched from your first paycheque and vested immediately.

How we hire

1. Apply 5 minutes

   Send links to what you have built and a few lines on why this role. No cover letter needed.

2. Intro call 30 minutes

   A conversation with the hiring manager about the work, the team and what you want next.

3. Work sample 2–4 hours

   A paid take-home or a pairing session on a real-shaped problem. You pick the format.

4. Studio day One day

   Meet the team, go deep on your work and ours. Remote or in person.

5. Offer Within 5 days

   A written offer with the full range, the equity and the reasoning behind both.