Human-AI Collaboration for Engineering Design Problems
Content
Speaker
Abstract
With the advent of ChatGPT and other chatbot products as well as open source LLMs, we can now truly use natural language to interact with computers in a way it was not possible before. The field of engineering design, due to its focus on search and optimization problems, has historically benefited a lot from classical AI techniques such as genetic algorithms, simulated annealing, and expert systems, but always with huge limitations due to the open-endedness of design problems. LLMs can bridge the gap between engineer thoughts and the optimization algorithms that can help create the best products and systems. This talk covers a lot of the recent work in the field, and the different ways in which LLMs and agentic workflows have been used in combination with classic optimization techniques for this field.
Bio
Dr. Antoni Viros i Martin is currently a Staff Research Scientist at IBM, and holds a PhD in Aerospace Engineering from Texas A&M University. His current research focuses on novel hardware-software optimization techniques for recent AI models like LLMs and State Space models. He is a contributor to Pytorch. Before working on optimization, his research focused on human-AI collaboration applied to the field of space systems design, with collaborations with both NASA and ESA, as well as DARPA and NSF.