Lead Research Scientist

Neo Research; United States; Remote; USD 180,000 - 240,000; Full time

Job summary

Effectief Altruisme seeks a lead research scientist to shape the organization\'s agenda on frontier-model risks and evaluations. You will set the research direction, lead major projects on alignment and model testing, and mentor researchers while upholding rigorous evidence standards.

You will build collaborations with AI safety institutes, frontier labs, and model developers, representing the research externally. A track record in AI safety research is required and a PhD is preferred.

Qualifications

• Track record in AI safety research.
• Experience leading large research projects.
• Experience collaborating with AI safety institutes and industry partners.

Responsibilities

• Set the research agenda and lead major projects on misalignment, loss of control, and model evaluation methodology.
• Mentor researchers, guide experimental design, and maintain rigorous evidence standards across the organization's work.
• Build collaborations with AI Safety Institutes, frontier labs, and model developers; represent research externally.

Skills

Research leadership

AI safety

Experiment design

Model evaluation

Education

PhD in a relevant field

Job description

Lead research scientist role shaping an AI safety organization's agenda on frontier-model risks and evaluations.

• Set Neo's research agenda and lead major projects on misalignment, loss of control, and model evaluation methodology.
• Mentor researchers, guide experimental design, and maintain rigorous evidence standards across the organization's work.
• Build collaborations with AI Safety Institutes, frontier labs, and model developers; represent research externally.
• Requires track record…

Als onderzoeksleider bepaal je welke vragen een heel team onderzoekt, en welke blijven liggen. Neo werkt samen met overheidsinstituten voor AI-veiligheid en met de bedrijven die de krachtigste modellen bouwen, dus de testmethoden die je hier ontwikkelt, kunnen ook bepalen hoe zij nagaan of een AI-systeem doet wat de makers bedoelen.