Conceptual Reasoning Researcher (Contract)

Trajectory Labs; Polska; Hybrid; PLN 434.000 - 814.000; Vollzeit

Zusammenfassung

Trajectory Labs is hiring researchers to design evaluations of conceptual reasoning in AI safety — tackling hard, open-ended problems in AI governance, alignment, macrostrategy, and economics, each paired with a rubric to distinguish real reasoning from filler. You’ll work 20–40+ hours weekly, remotely or at Constellation in Berkeley.

As an early-stage startup, you should expect responsibilities to evolve quickly as priorities shift.

Qualifikationen

• Domain depth in AI risk, alignment, macrostrategy, economics, or governance.
• Articulable taste: explain why one argument is better and encode in rubrics.
• Calibration under pushback: revise judgments when challenged and defend them.
• Agent-native research: fluent with frontier models and agentic tools like Claude Code.

Aufgaben

• Find hard conceptual questions in AI safety — alignment, macrostrategy, economics, governance — where frontier models still reason badly.
• Turn each into an evaluation task: a well-posed problem and an opinionated rubric that rewards excellent reasoning.
• Run your tasks against frontier models and iterate until the scores separate strong answers from plausible-sounding weak ones.
• Give and take detailed QA — tasks ship after review, not on first draft.

Kenntnisse

Domain depth

Articulable taste

Calibration under pushback

Agent-native research

Ausbildung

PhD-level research experience

Jobbeschreibung

• Engagement: Hourly contract, 20–40+ hours/week

• Start: ASAP

• Contract length: ~3 months, with potential to renew

• Reports to: Ryan Carey (MTS, Head of Conceptual Reasoning)

• Location: Remote, or on-site at Constellation in Berkeley

• Compensation: $80–150/hr

Overview

Trajectory Labs is hiring researchers to design evaluations of conceptual reasoning in AI safety — hard, open-ended problems in AI governance, alignment, macrostrategy, and economics, each paired with a rubric that separates real reasoning from plausible-sounding filler. You'll work 20–40+ hours a week, remotely or alongside us at Constellation in Berkeley.

We're an early-stage startup, so you should expect and enjoy that your responsibilities will grow and priorities change quickly.

About us

Our mission is to automate AI safety, to pave the way for a future where the vast majority of AI safety work is done by AI models.

To that end, we build safety and alignment evals, red-teaming programs, and RL environments for frontier AI labs. Our areas of focus include red teaming, control, anti-distillation, AI safety reasoning, reward-hacking propensity evaluations, and embedded assessments.

Although Trajectory is only a year old, we have already been featured in Anthropic's Claude Opus 5.5 System Card (the only commercial red team to jailbreak its cyber safeguards) and Google DeepMind's Gemini 4 Argon Safety & Security Report, led Anthropic's prompt injection evaluation of Auto mode, and been featured in Meta's Muse Spark Safety & Preparedness Report.

About the role

Over time, it is important that we can spend money to get useful AI safety work out of AI models, and this requires them to get better at some hard-to-supervise fuzzy tasks in AI safety and related domains. We're partnering with frontier AI labs to close that gap.

We’re hiring researchers with specific expertise in AI governance (including law, international relations or technical governance), alignment, macrostrategy, or economics, who are capable of evaluating novel arguments in these areas and devising rubrics that can separate excellent reasoning from plausible-sounding filler. The tasks you produce become part of the benchmarks that frontier labs use to measure — and improve — how their models reason about making TAI go well.

It's important to us that our work is robustly positive, so we focus our evals on the capabilities that differentially accelerate safety judgment, not capabilities in general. While there will be some spillovers, we think it is important that AI models are able to help with safety and alignment, and that this outweighs the second-order effects. You can help us with this goal by pointing us to evidence and reasons we might be missing in making this decision and designing tasks that differentially advance safety judgment.

What you'll do

• Find hard conceptual questions in AI safety — alignment, macrostrategy, economics, governance — where frontier models still reason badly

• Turn each into an evaluation task: a well-posed problem and an opinionated rubric that rewards excellent reasoning and penalizes slop

• Run your tasks against frontier models and iterate until the scores separate strong answers from plausible-sounding weak ones

• Give and take detailed QA — tasks ship after review, not on first draft

What we're looking for

We're looking for researchers in any of the four areas below, or adjacent fields, who have strong, defensible opinions about what good reasoning looks like in their specific area of expertise.

• Governance: compute governance and verification, international agreements and security, AI law and institutions

• AI safety & alignment: alignment theory, AI control and oversight, evaluations, model psychology and AI welfare

• Philosophy & macrostrategy: ethics, AI moral status, takeoff dynamics, power concentration and long-run trajectories

• Economics & decision theory: growth theory, AI’s effects on labor and capital, game theory and mechanism design, decision theory

Essential

• Domain depth: you have produced original analysis on AI risk, alignment, macrostrategy, economics, or governance that you stand behind.

• Articulable taste: you can say why one argument in your field is better than another, precisely enough to encode it in a rubric.

• Calibration under pushback: you revise your judgment when the argument against it is good, and defend it with reasons when it isn't.

• Agent-native research: you work fluently with frontier models and agentic tools — you get real leverage from Claude Code or similar in your own work, and can tell the difference between output that is correct and output that merely sounds right.

Desirable

• PhD-level research experience. All of our CR staff have studied and/or worked at top universities, but a PhD is not a hard requirement.

• Experience writing grading rubrics, exam questions, or evaluation criteria

• Peer-reviewed or widely-cited work in one of the domains above

• Experience using frontier models to answer messy questions in your own research

Role highlights

Your taste in arguments gets encoded into evals frontier labs train and measure against, the feedback loop is short — a task goes from draft to scored model runs in days — and you set your own hours.

Work with our top research team including PhDs or research fellows from Oxford University (x4) and LSE (x2) and the leading figures in your field.

Other details:

• Remote-friendly, async by default; desk space available at Constellation in Berkeley

• 20–40+ hours/week, flexible as your availability changes

• Starts immediately, rolling review