AI Safety Researcher: Multi-Agent & Evaluation

CTI Clinical Trial and Consulting Services; Roma; In loco; EUR 60.000 - 90.000; Tempo pieno

Descrizione del lavoro

Icaro Foundation in Rome seeks a mid- to senior-level researcher to advance safety research for frontier AI systems. You will review literature, design experiments, implement code, and contribute to papers and technical reports.

The role blends research and engineering: write Python code, extend tools, document configurations, and share reproducible results. Remote options may be considered for Europe or China, with in-person collaboration in Rome preferred.

Competenze

• Experience with LLM evaluations, red-teaming, or benchmark development.
• Strong technical or quantitative background from university study or projects.
• Strong Python skills and familiarity with Git and debugging.
• Experience with ML or LLMs through at least one project or research contribution.
• Understanding of experimental reasoning, statistics, and uncertainty.
• Ability to read technical papers and explain methods in English.
• Strong interest in AI safety and urgency to reduce AI risks.
• Willingness to revise views based on evidence.

Mansioni

• Review literature, compare methods, and identify questions worth investigating.
• Turn research questions into experimental protocols with baselines and evaluation criteria.
• Implement and run experiments with frontier and open-weight models.
• Analyze results and model traces, assess confounders and alternative explanations.
• Contribute to research papers, benchmarks, and technical reports.
• Develop the research codebase with Python, extend tools, and fix bugs.
• Add tests and reproducible configurations for inspectability.
• Document experiments, findings, and limitations.

Conoscenze

LLM evaluations

Red-teaming

Benchmark development

Python programming

Experimental design

Statistics basics

Critical reading

AI safety interest

Problem solving

Strumenti

Git

Descrizione del lavoro

Icaro Foundation in Rome seeks a mid- to senior-level researcher to advance safety research for frontier AI systems. You will review literature, design experiments, implement code, and contribute to papers and technical reports.

The role blends research and engineering: write Python code, extend tools, document configurations, and share reproducible results. Remote options may be considered for Europe or China, with in-person collaboration in Rome preferred.