Kavosh Asadi

WORK HISTORY

EDUCATION

SKILLS

ABOUT KAVOSH ASADI

At Amazon, I explore core principles underlying goal-based decision making in sequential environments. I often model these problems using reinforcement learning. I am particularly interested in looking at the reinforcement learning problem through the lens of optimization and in complicated domains such as those that arise when using powerful deep neural networks for function approximation. In my view, a really intriguing application of AI is that of conversational agents. During my PhD, I did internships at Microsoft Research where I worked on and published in the area of developing conversational agents using supervised and reinforcement learning. The emergence of large language models has drawn my attention into this area again, where I am currently focused on supervised fine-tuning of LLMs for specific tasks as well as reinforcement learning from human feedback. I am passionate about AI education, and strive to reduce the barrier of entry to the field for junior AI students. To this end, I am contributing to the effort to add a chapter on reinforcement learning to the D2L book: https://d2l.ai. Before Amazon, I got my PhD at Brown University where I worked with my amazing mentor Michael Littman. Prior to Brown, I did my Master's at the University of Alberta where my AI foundation was heavily influenced by Rich Sutton. I have recently become obsessed with the violin, and after a year of hard work, have learned to play the main title of Game of Thrones from Ramin Djawadi. I am also exploring projects in the intersection of machine learning and music. I am a permanent resident (green-card holder) of the United States.