Apply on Employer Site

Inside Higher Ed · 4 days ago

Assistant Research Scientist AI Safety Center for Human-Compatible Artificial Intelligence

Berkeley, CA

Full-time

Onsite

New Grad, Entry Level

$124K/yr - $156K/yr

The University of California, Berkeley is seeking an Assistant Research Scientist to conduct and lead research in the domain of AI safety and well-founded AI. The role involves developing decision-making models, authoring peer-reviewed articles, mentoring PhD students, and facilitating collaborations within and beyond UC Berkeley.

Digital MediaEducationHigher EducationJournalismRecruiting

No H1B

Responsibilities

The candidate will conduct and lead research on AI safety and well-founded AI, including: reinforcement learning, AI planning, state abstraction, action abstraction, world modeling, decision-making under partial observability, and methods for analyzing and interpreting learned representations

Research will include developing novel decision-making models and algorithms with strong theoretical foundations, conducting rigorous empirical testing and evaluation of these models and algorithms, and establishing formal frameworks for safe and aligned AI systems

The candidate will author peer-reviewed articles, lead multiple research projects, mentor PhD students and other junior researchers, set research directions, write technical papers and reports, and present at research conferences and academic labs

The candidate will facilitate collaborations both within and beyond UC Berkeley

Qualification

Reinforcement learningDecision-making under uncertaintyPythonMachine learning librariesPhD in computer sciencePublication recordExperience mentoringFluency in non-technical disciplineCreativityProblem-solving

Required

PhD or equivalent international degree

Preferred

PhD in computer science, mathematics, statistics, or related field

Expertise in reinforcement learning, planning, and decision-making under uncertainty

Experience with one or more of: state abstraction, temporal abstraction, model-based methods, partial observability, or interpretability of learned representations

Fluency in at least one non-technical discipline such as cognitive science, philosophy, social science, or policy development

Strong publication record at top-tier conferences and journals

Experience conducting and leading full-time research

Experience organizing and reviewing for conferences and workshops

Demonstrated creativity and problem-solving skills

Experience mentoring graduate students or other junior researchers

Proficiency in Python and/or C++ for implementing and testing algorithms

Experience with PyTorch, JAX, NumPy, SciPy, and other machine learning libraries

Experience developing open-source research software or AI benchmarks