About Me
I am a recent Ph.D. graduate from Carnegie Mellon University advised by Jonathan Cagan and Chris McComb. I’m interested in advancing real-time AI collaboration in both creative and technical problem-solving. I’m currently exploring full-time opportunities in machine intelligence, behavior learning, and human-machine collaboration.
My thesis proposes several agent frameworks that enable representation learning, effective exploration, and policy generalization in the generative design domain. My research enables learning from scratch in arbitrarily complex actions using novel deep learning and search methods.
Prior to my Ph.D. I graduated from the Indian Institute of Technology Jodhpur in 2017 with a Bachelor of Technology in Mechanical Engineering. I was advised by Kaushal Desai, Suril V. Shah for my undergraduate thesis.
Apart from research, I love to go on hikes and click pictures whenever I can. I also enjoy playing competitive multiplayer games and learning new skills online.
Research Interests
Machine Intelligence in Design
I am interested in building agent frameworks that solve complex decision-making problems. Design problems constitute several unique challenges with large and unstructured state-action spaces, state representations, and creativity introducing an interesting benchmark for intelligent agent development.
Human-Machine Collaboration
With machines surpassing human skill levels in several problem-solving processes, a collaboration between humans and machines is imminent. I am interested in exploring real-time collaboration mechanisms between humans and machines to augment human decision-making.
Behavior Learning and Interpretability
High-performing problem solvers have diverse and complex decision-making behavior. I am interested in understanding the underlying dynamics of that behavior and developing quantitative explanations for such behavior, which enables the transfer, learning, and interpretability of strategies.
Major challenges in machine intelligence in design
The challenges of representation learning, novel discovery or effective exploration, and policy generalization or transfer learning are addressed in my dissertation research
Recent Highlights
Self Learning Design Agent (SLDA): Enabling Deep Learning and Tree Search in Complex Action Spaces
Ayush Raina, Jonathan Cagan, Chris McComb
Journal of Mechanical Design 2022 (under review)
Design Strategy Network: A deep hierarchical framework to represent generative design strategies in complex action spaces.
Ayush Raina, Jonathan Cagan, Christopher McComb
Journal of Mechanical Design, 144(2):1–36, 10 2021
©Ayush Raina, 2022
All Rights Reserved
Linked Pages:5