Autonomous AI Safety Research Scientist

White Circle; Paris; Hybrid; EUR 120.000 - 190.000; Vollzeit

Benefits dieser Stelle

Equity

Flexible hybrid setup

Relocation support to Paris after prob

Premium private health insurance

Mental health support

Lunch and dinner when at the office

Learning and development support

All hardware/tools needed

Team off-sites twice a year

Zusammenfassung

White Circle, a safety-focused AI company, seeks a research scientist to push LLM agents to failure and study breakdowns. You will craft adversarial, large-scale environments and orchestrate many agents to observe failures arising from agent behavior, not external scripts.

You will run end-to-end experiments, publish findings, and help model agent behavior while contributing to safety benchmarks and writeups. A PhD or MSc in a quantitative field is valued.

Qualifikationen

• Built non-trivial agent environments or end-to-end research pipelines.
• Strong software/AI engineering skills with parallel orchestration.
• Track record in empirical agent research, red-teaming, or post-training.
• Fast empirical iterator; can define questions without a playbook.
• Fluent with frontier models and coding agents daily.

Aufgaben

• Build adversarial and complex agent environments near the edge of capability.
• Create realistic multi-agent setups and observe emergent failures.
• Run end-to-end experiments against external APIs and internal models.
• Catalogue agent failure modes and surface them at scale.
• Turn findings into internal models and public writeups.

Kenntnisse

Agent environments

Multi-agent orchestration

Experiment design

Empirical research

Frontier models

Ausbildung

MSc or PhD in ML/CS/Cognitive Science/Computational Neuroscience/Physics

Jobbeschreibung

White Circle, a safety-focused AI company, seeks a research scientist to push LLM agents to failure and study breakdowns. You will craft adversarial, large-scale environments and orchestrate many agents to observe failures arising from agent behavior, not external scripts.

You will run end-to-end experiments, publish findings, and help model agent behavior while contributing to safety benchmarks and writeups. A PhD or MSc in a quantitative field is valued.