Building Scientific Foundation Models: Challenges, Methodologies, and Semiconductor Manufacturing Applications

Noseong Park
13:10 - 13:30
, June 8
June 8

Scientific foundation models aim to do for partial-differential equations what language models have done for text: provide a single, reusable network that can generalize across problem classes with no task-specific fine-tuning. Achieving this vision poses unique obstacles — from curating heterogeneous training corpora and encoding physical constraints to designing data-efficient architectures and scaling training on modern accelerators. In this talk I will: i) Survey the emerging literature on physics-informed and operator-learning frameworks; ii) Dissect the technical hurdles that separate today’s bespoke surrogates from tomorrow’s foundation models, including data scarcity in high-fidelity simulations, conditioning across disparate boundary conditions, and robustness to extrapolation; iii) Present our lab’s recent progress toward a single model that solves a wide spectrum of fluid-dynamics PDEs; iv) Outline a research roadmap that the community can pursue to realize practical scientific foundation models for semiconductor applications.

About the speaker

Noseong Park

Noseong Park

Noseong Park is currently a tenured associate professor in the School of Computing and the director of the Big Data Analytics and Learning Laboratory at the Korea Advanced Institute of Science and Technology (KAIST) in South Korea. Before joining KAIST, he was a tenured associate professor at Yonsei University from 2020 to 2024. Prior to that, he served as an assistant professor at George Mason University from 2018 to 2019 and at the University of North Carolina at Charlotte from 2016 to 2018. He received his Ph.D. in Computer Science in 2016 from the University of Maryland, College Park, under the supervision of Professor V. S. Subrahmanian. He earned his Master’s degree from KAIST and his Bachelor’s degree from Soongsil University. He has extensive experience in data mining and machine learning, with applications in scientific machine learning, deep generative learning, spatiotemporal processing, time series processing, recommender systems, and more. He has published over 70 papers in top-tier venues, including NeurIPS, ICLR, ICML, KDD, WWW, VLDB, ICDM, WSDM, SIGIR, IJCAI, and AAAI. Notably, his VLDB paper introducing TableGAN, a method for tabular data synthesis, is the most cited VLDB paper published in the last six years. He has received several accolades, including the Best Student Paper Award as an advisor from IEEE BigData 2022, the Best Demonstration Runner-Up Award from IJCAI 2019, and the Best Paper Runner-Up Award from ASONAM 2016. Additionally, he was honored with multiple Samsung Humantech Paper Awards, one of the most prestigious awards in Korea, in 2022 and 2023. He has secured over $530K in research grants as a principal investigator or co-principal investigator from the NSF, the Office of Naval Research, and other sources during his time in the US. Including grants received in Korea since 2020, his total research funding amounts to $3 million. In particular, his recent research grant on the scientific foundation model for solving continuous equations had been selected by Samsung Science & Technology Foundation. He has also served (or is currently serving) as an area chair or a program committee member for many top-tier conferences, including ICML, ICLR, NeurIPS, KDD, WWW, AAAI, ICDM, and ICWSM. He will serve as a general chair for CIKM, a top-tier venue for recommender systems and information retrieval, in 2025.

Noseong Park is currently a tenured associate professor in the School of Computing and the director of the Big Data Analytics and Learning Laboratory at the Korea Advanced Institute of Science and Technology (KAIST) in South Korea. Before joining KAIST, he was a tenured associate professor at Yonsei University from 2020 to 2024. Prior to that, he served as an assistant professor at George Mason University from 2018 to 2019 and at the University of North Carolina at Charlotte from 2016 to 2018. He received his Ph.D. in Computer Science in 2016 from the University of Maryland, College Park, under the supervision of Professor V. S. Subrahmanian. He earned his Master’s degree from KAIST and his Bachelor’s degree from Soongsil University. He has extensive experience in data mining and machine learning, with applications in scientific machine learning, deep generative learning, spatiotemporal processing, time series processing, recommender systems, and more. He has published over 70 papers in top-tier venues, including NeurIPS, ICLR, ICML, KDD, WWW, VLDB, ICDM, WSDM, SIGIR, IJCAI, and AAAI. Notably, his VLDB paper introducing TableGAN, a method for tabular data synthesis, is the most cited VLDB paper published in the last six years. He has received several accolades, including the Best Student Paper Award as an advisor from IEEE BigData 2022, the Best Demonstration Runner-Up Award from IJCAI 2019, and the Best Paper Runner-Up Award from ASONAM 2016. Additionally, he was honored with multiple Samsung Humantech Paper Awards, one of the most prestigious awards in Korea, in 2022 and 2023. He has secured over $530K in research grants as a principal investigator or co-principal investigator from the NSF, the Office of Naval Research, and other sources during his time in the US. Including grants received in Korea since 2020, his total research funding amounts to $3 million. In particular, his recent research grant on the scientific foundation model for solving continuous equations had been selected by Samsung Science & Technology Foundation. He has also served (or is currently serving) as an area chair or a program committee member for many top-tier conferences, including ICML, ICLR, NeurIPS, KDD, WWW, AAAI, ICDM, and ICWSM. He will serve as a general chair for CIKM, a top-tier venue for recommender systems and information retrieval, in 2025.

ACME Conference logo
© 2025 VLSI Workshop Science Meets AI
ACME Conference logo
© 2025 VLSI Workshop Science Meets AI
ACME Conference logo
© 2025 VLSI Workshop Science Meets AI