Speakers

His research goal is to create AI algorithms that can solve any pathfinding problem. To accomplish this, his research group builds deep reinforcement learning, heuristic search, formal logic, and large language models. He led the creation of DeepCubeA, an AI algorithm capable of solving puzzles such as the Rubik’s cube without human guidance and currently applies his research to fields such as chemical synthesis, quantum computing, robotics, and education. He earned his Ph.D. in computer science from the University of California, Irvine

The objective of pathfinding is to find a sequence of actions that forms a path from a start state to a goal state. While the specification of a goal (i.e., the set of goal states) is fundamental to pathfinding, it is often the case that only the high-level properties a goal state must have can be specified, yet no state that has these properties is known. For example, in chemical synthesis, a practitioner may know the binding affinities and solubility of the desired molecule but not know of any molecules that have these properties. Being able to specify and reach high-level goals would not only allow for the synthesis of such molecules but also lead to the discovery of novel molecules. I will describe our approach of using formal logic to specify high-level goals and deep reinforcement learning to learn heuristic functions to reach highlevel goals with a heuristic search. I will discuss our current work, made possible by the NSF CAREER program, on generalizing this approach to more expressive goal specifications and on tightening the integration of deep reinforcement learning and search. 

Forest Agostinelli, Ph.D.
Assistant Professor
University of South Carolina, USA

Monday, November 9th, 2026
10:00 a.m. – 11:00 p.m.
USF LIBRARY-6th Floor, Room 645

Read Forest Agostinelli's Abstract and Bio 

Solving Pathfinding Problems with High-Level Goal Specifications

Trustworthy Multimodal Learning Systems and their Applications

Khoa Luu, Ph.D.
Department of Electrical Engineering and Computer Science
University of Arkansas, USA

Read Khoa Luu's Abstract and Bio 

---

Towards Extensible and Reproducible Visual Analytics: AI-Guided and Human-Grounded Frameworks for Collaborative System Authoring

Fabio Miranda, PhD
University of Illinois Chicago
Chicago, USA

Read Fabio Miranda's Abstract and Bio 

---

From Data to Knowledge to Better Clinical Outcomes: Leveraging Real-World Data to Assess Health Across the Lifespan

Mamoun Mardini
Assistant Professor in the Department of Health Outcomes and Biomedical Informatics at the University of Florida

Read Mamoun Mardini's Abstract and Bio 

---

Cooperation Algorithms in Multi-Agent Reinforcement Learning

Reza Azadeh
Associate Professor at the Miner School of Computer and Information Sciences at the University of Massachusetts Lowell

Read Reza Azadeh's Abstract and Bio 

---

Designs to Support Better Visual Data Communication

Cindy Xiong Bearfield
Assistant Professor at School of Interactive Computing at Georgia Institute of Technology

Read Cindy Bearfield's Abstract and Bio 

---

AI-Driven Biometric Security and Healthcare Applications: From Anti-Spoofing to Adversarial Resilience

Nima Karimian
Assistant Professor at the Lane Department of Computer Science and Electrical Engineering, West Virginia University

Read Nima Karimian’s Abstract and Bio 

---

AI-Driven Biometric Security and Healthcare Applications: From Anti-Spoofing to Adversarial Resilience

Austin Downey
Associate Professor in Mechanical, Aerospace, and Civil Engineering at the University of South Carolina

Read Austin Downey's Abstract and Bio 

---

Accelerating AI at the Edge: Innovations in Hardware and Algorithms

Ramtin Zand
Assistant Professor of Computer Science at the University of South Carolina

Read Ramtin Zand's Abstract and Bio 

---