Research Scientist/Engineer (Science of Scheming)

TL;DR

Research Scientist/Engineer (AI/ML): Building a new hard science focused on understanding and predicting AI scheming risks with an accent on RL dynamics, scaling laws, and novel evaluation techniques. Focus on empirical foundations, AI cognition, and resolving issues in LLM training.

Company

hirify.global is building a new hard science to understand and predict AI scheming risks.

What you will do

• Collaborate with leading AI developers and partner labs to directly impact how capable AI systems are built and deployed.
• Deeply study RL dynamics that lead to the emergence of reward-seeking or misaligned preferences in AI.
• Build empirical foundations for "Scaling laws of scheming" to predict how scheming risks evolve.
• Develop novel and ambitious evaluation techniques for highly evaluation-aware models.
• Deep dive into AI cognition to discover reasoning patterns in frontier AI systems.

Requirements

• Fast-paced empirical research: Ability to design and execute experiments, speeding up iteration cycles.
• Conceptual insights about scheming: Deep understanding of AI scheming problems and relevant literature.
• Strong software engineering skills, particularly in Python.
• Intense interest in AI progress: Stay updated on model releases and actively understand AI cognition.
• Experience with RL-training LLMs, including debugging learning instabilities.
• Strong analytical skills: Rigorous quantitative skills from fields like scaling laws, statistical physics, dynamical systems, or applied statistics.

Culture & Benefits

• Opportunities to collaborate with leading AI developers and partner labs.
• Direct impact on how the most capable AI systems are built and deployed.
• Open to candidates from diverse backgrounds, including self-taught individuals.
• Focus on pioneering a new hard science from the ground up.

Hiring process

• Interviews are actively conducted, aiming to fill the role as soon as a suitable candidate is found.