Augustine Twumasi
He/Him/His
Assistant Professor
Mathematics, Statistics and Computer Science
University of Wisconsin-Stout
Biography
Augustine Twumasi is an Assistant Professor of Artificial Intelligence & Machine Learning in UW–Stout’s Department of Mathematics, Statistics, and Computer Science. His research tackles a wide range of science and engineering problems by combining scientific machine learning, advanced numerical methods, optimization, and uncertainty quantification to create data-efficient, physics-informed AI models for complex dynamical systems. He is also interested in computer-vision-driven quality control and scalable MLOps for industrial workflows. His work has been shaped by research appointments at Oak Ridge National Laboratory and Pacific Northwest National Laboratory.
Degrees Earned
Ph.D. Computational Science - 2025 MS Computational Science - 2024 MS Mathematics - 2021 BSc. Mathematics -2018
Research Areas
Applied Mathematics; Data Science; Engineering; Machine Learning/AI
Research Interests
My research lies at the intersection of scientific machine learning and applied mathematics, focusing on developing physics-informed, uncertainty-aware AI models for complex dynamical systems. I apply these methods to robotics and additive manufacturing, advancing data-efficient approaches for process optimization, defect prediction, and autonomous system control.
Topical Areas
Applied Computer Science; Applied Mathematics; Artificial Intelligence and Intelligent Systems; Informatics, Analytics and Information Science; Materials Engineering
Research Synergy
My research provides scientific machine learning tools that integrate physics-based modeling, optimization, and uncertainty quantification to enhance the understanding and control of complex engineering systems. These approaches directly support advances in additive manufacturing, robotics, and materials design by improving predictive accuracy, process efficiency, and autonomous decision-making across experimental and computational domains.
Motivation
I first learned about the SRP program in September 2022 when I was selected for the Broader Engagement Program at the SIAM Conference on Mathematics of Data Science (BE@SIAM22) in San Diego. As a participant in the program during the summer of 2023, I gained invaluable collaborative experience that helped shape my research thrust. At that time, I was still defining my research focus, but my subsequent three-month internship at Oak Ridge National Laboratory made possible through SRP, which directly led to the development of my Ph.D. dissertation. The program has been instrumental in my growth as a researcher and in my current role as an Assistant Professor of AI and Machine Learning. I continue to advocate for SRP among my students, as I believe its collaborative model demonstrates how meaningful partnerships drive innovation. Participating again would further strengthen my work as an early-career researcher and allow me to mentor students in the same spirit of collaboration that shaped my own path.
Supervising Students Plan
I plan to structure the supervision of my student team as a collaborative, research-focused mentorship centered on developing reinforcement learning algorithms for adaptive time stepping with applications in robot path planning. My student will begin with training in RL fundamentals, numerical integration techniques, and simulation frameworks before implementing adaptive control schemes for dynamic robotic systems. We will hold weekly meetings to analyze algorithm performance, stability, and efficiency while refining model architectures and reward functions. Student will participate in paper writing, documentation, and conference presentations to strengthen both their research and communication skills. My role will be to provide technical guidance, integrate their work into ongoing research in scientific machine learning, and mentor them toward independent, high-impact research outcomes.
Student Merit
I worked with Patrick during my Ph.D. when he was a master’s student in our lab, and I am currently serving as an external committee member for his Ph.D. research. In the summer of 2025, we collaborated on developing a numerical scheme for adaptive time stepping in robot path planning, where he demonstrated strong analytical skills and consistent progress. Patrick is highly adaptable, quickly mastering new computational techniques and integrating feedback effectively. With his solid background in mechanical engineering and growing expertise in scientific machine learning, he is well-prepared to apply these methods to robotics and advanced control problems.
Lightning Talk Title
Physics-Guided Machine Learning for Phase-Field Stability and Microstructure Control
Keywords
Phase-Field Modeling Additive Manufacturing Machine Learning Deep Reinforcement Learning Path Planning