About the job
About the Institute of Foundation Models
At the Institute of Foundation Models, we are committed to pioneering research that enhances the understanding, utilization, and risk management of foundation models. Our mission focuses on advancing research, cultivating future AI innovators, and making transformative contributions to a knowledge-driven economy.
Join our dynamic team and engage in the forefront of groundbreaking foundation model training alongside leading researchers, data scientists, and engineers. You will address the most critical and impactful challenges in AI development, contributing to innovative AI solutions with the potential to revolutionize industries. Your strategic and creative problem-solving abilities will be key in positioning MBZUAI as a premier global center for high-performance computing in deep learning, fostering discoveries that will inspire the next generation of AI visionaries.
Position Summary
As a valued member of the Agents team, you will confront research and engineering obstacles related to training advanced agentic language models capable of employing reasoning and tool use to perform tasks on computers. The scope of challenges includes constructing high-quality training datasets, developing robust evaluation methods, and innovating reinforcement learning infrastructures, algorithms, and environments. We are seeking candidates with diverse expertise, ranging from algorithmic and empirical research to large-scale systems and optimization, reflecting the wide array of compelling problems we aim to address. Depth in any of these areas is highly regarded; there is no single “correct” background.
Key Responsibilities
· Propose, prototype, and scale systems and algorithms that support agentic learning.
· Develop innovative paradigms for agentic behavior that extend beyond existing methodologies.
· Contribute to technical reports, research publications, and open-source software initiatives.
· Collaborate with interdisciplinary teams to create state-of-the-art foundation models.
