Search
Go
Frontier AI Safety Evaluator — Risk & Alignment
Summary by AIEngineer.careers
Frontier AI Safety Evaluator — Risk & Alignment is a AI Safety Engineer role (full-time). with Obsidian. in MADRID, ES. Compensation shown: $75K–$119K (est.). Imported listing (source: empleo24.es). Apply on the employer's site (empleo24.es).
Imported listing (source: empleo24.es) · Apply on empleo24.es
MADRID · ES
$75K–$119K (est.)
Full-time
MID
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 empleo24.es
This role focuses on evaluating the safety, quality, and alignment of frontier AI models, specifically addressing policy-sensitive topics. The candidate will collaborate with researchers to identify unsafe model outputs and refine evaluation rubrics for Reinforcement Learning from Human Feedback (RLHF) and Supervised Fine-Tuning (SFT).
Responsibilities
- Evaluate safety and alignment of frontier AI models across various domains.
- Identify and categorize unsafe or biased model outputs.
- Collaborate with research teams to refine RLHF/SFT rubrics.
- Analyze policy-sensitive topics to ensure model compliance.
Requirements
- Experience in AI model evaluation or safety research.
- Understanding of RLHF and SFT processes.
- Strong analytical skills for identifying alignment issues.
- Ability to collaborate effectively with engineering and research teams.
Nice to Have
- Background in ethics, philosophy, or social sciences related to AI.
- Prior experience with large language model training pipelines.
- Familiarity with automated safety testing tools.
Obsidian
Toronto · CA · 12+ employees
Obsidian (developed by Dynalist Inc.) is a powerful and extensible note-taking application designed for personal knowledge management. It operates on a local vault of Markdown files, emphasizing privacy, customizability through plugins and themes, and long-term data ownership.
