Search
Go
Research Scientist (AI Behaviours)
Summary by AIEngineer.careers
Research Scientist (AI Behaviours) is a AI Safety Engineer role (full-time). with Whitecircle. in BERLIN, DE. Imported listing (source: clever.cv). Apply on the employer's site (clever.cv).
Imported listing (source: clever.cv) · Apply on clever.cv
BERLIN · DE
—
Full-time
SENIOR
The sections below reproduce the third-party job description for reference. AIEngineer.careers does not write or control this text.
Imported job description
Sourced from clever.cv
White Circle is seeking a Senior Research Scientist to investigate AI safety and alignment issues, specifically focusing on how LLM agents fail or exhibit unsafe behaviors in real-world scenarios. You will design empirical experiments, build automated audit tools, and pressure-test frontier models to identify misalignment and bias before they reach production. This role bridges deep research with strong ML engineering capabilities to improve AI reliability.
Responsibilities
- Own end-to-end research projects from hypothesis to falsifiable experiments and published results.
- Develop automated audit agents to discover and characterize suspect model behaviors at scale.
- Study misalignment and bias emergence during user-agent interactions to create shippable evaluations.
- Conduct white-box and black-box investigations to understand model failure modes in high-stakes scenarios.
- Publish findings as conference papers and blog posts while feeding insights into internal guardrails.
Requirements
- Track record of empirical research in agent behavior, model evaluation, or alignment.
- Strong ML engineering skills: ability to independently build research MVPs involving fine-tuning and evals.
- Expertise in experimental design under real conditions, isolating failure modes, and calibrating baselines.
Whitecircle
Paris · FR · 25+ employees
White Circle is an AI security and control platform that helps enterprises test, monitor, and protect AI models in production. It offers solutions for blocking unsafe inputs, preventing jailbreaks, and optimizing performance.
