Workshop 1
Innovating Semiconductor Manufacturing: Science Meets AI
Organizer : Jaehoon Jeong (Samsung Electronics)
Workshop Special website
Further details of this workshop :https://sciencemeetsai.com/.
Purpose of the Workshop
The primary goal of this workshop is to explore how science-based artificial intelligence (AI) can address the unique challenges of semiconductor manufacturing, an industry that demands both precision and innovation. By integrating fundamental scientific principles with advanced AI technologies, the workshop aims to:
- Facilitating Knowledge Exchange: Provide a platform to share breakthroughs in science-based AI applications for semiconductor processes.
- Promoting Cross-Disciplinary Collaboration: Bridge the gap between semiconductor researchers, AI scientists, and industry practitioners to address critical challenges.
- Fostering Science-Based Digital Twin: Highlight how AI grounded in scientific methodologies can optimize manufacturing processes, accelerate material discovery, and enhance manufacturing efficiency.
Why This Workshop is Essential?
- Rising Complexity in Semiconductor Manufacturing: As chip designs become smaller, more powerful, and more complex, traditional manufacturing processes face limitations. Science-based AI, informed by physics, chemistry, and material science, offers targeted solutions to these challenges by enabling more accurate modeling, prediction, and optimization.
- Industry Demand for AI Integration: The semiconductor industry is increasingly adopting AI to stay competitive. However, there is a need for a structured platform to discuss best practices, share case studies, and address barriers to adoption.
- Limitations of General AI Models: General-purpose AI models often lack the depth of understanding required for specialized applications like semiconductor manufacturing. When applied to complex problems with limited data, these models can produce results that violate physical laws or fail to account for critical scientific constraints. On the other hand, science-based AI mitigates these risks by embedding physical laws, chemical properties, and domain-specific knowledge into AI algorithms, ensuring accuracy and reliability.
- Accelerating Innovation: The convergence of AI and semiconductor manufacturing has the potential to revolutionize the industry, from advanced chip design to smart factories. Hosting this workshop can catalyze innovation by connecting stakeholders and fostering new ideas.
- Educational Impact: This workshop will provide valuable learning opportunities for participants, helping them stay at the forefront of technological advancements and preparing them for future challenges.
Proposed Topics
- An exploration of the necessity for science-based AI integration.
- Strategies for efficiently instructing AI in scientific disciplines
- Illustrative applications of science-based AI in the semiconductor manufacturing
The first topic was selected owing to Professor Jeong's profound expertise and prolific publication record, which provide rich insights into this area. The second topic is inspired by the unparalleled authority of Professor Anima, who stands as the foremost expert in the AI domain on this subject. For the third topic, we have enlisted a distinguished executive from a leading semiconductor company to elucidate the practical applications and transformative impact of science-based AI in the industry.
By implementing this structured operational plan, the workshop aims to provide a meaningful, impactful, and well-organized experience for all participants, fostering innovation and collaboration in the field of semiconductor manufacturing and AI.
About Organizer
Jaehoon Jeong is a seasoned professional with over 17 years of experience in digital twin technology. He currently serves as a Project Leader at Samsung Electronics, where he has been driving innovation and advancing technological solutions since joining the company in 2007. Dr. Jeong earned his Ph.D. in Electronic Engineering from Texas A&M University in 2006. His career at Samsung began with PKG and PCB development, which later expanded to include transistor modeling and 3-D IC modeling. These experiences laid the foundation for his expertise in bridging cutting-edge technology with real-world applications. Dr. Jeong’s recent focus is on introducing science-based digital twins, a transformative approach that combines scientific principles with artificial intelligence to enhance predictive modeling and operational efficiency. His innovative work in this area aims to redefine the integration of science and AI in industrial and technological contexts. Throughout his career, Dr. Jeong has been recognized for his contributions to the field, leading projects that bridge the gap between cutting-edge research and practical application. His work continues to shape the future of digital transformation in the tech industry.
- 1. Opening (Science Meets AI), Jaehoon Jeong, Project Leader, Simulation Solution, Samsung Electronics
- 2. An Exploration of the Necessity for Science-driven AI Integration, Animashree Anandkumar, Bren Professor at Caltech
-
Abstract Language models have been used for generating new ideas and hypotheses in scientific domains. For instance, language models could suggest new drugs or engineering designs. However, this is not sufficient to attack the hard part of science which is the physical experiments needed to validate the proposed ideas. This is because language models lack physical validity and the ability to internally simulate the processes. Traditional simulation methods are too slow and infeasible for complex processes observed in many scientific domains. We propose AI-based simulation methods that are 4-5 orders of magnitude faster and cheaper than traditional simulations. They are based on Neural Operators which learn mappings between function spaces.
- 3. Building Scientific Foundation Models: Challenges, Methodologies, and Semiconductor Manufacturing Applications, Noseong Park, Associate Professor at KAIST
-
Abstract 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.
- 4. Applications of Science-driven AI in the Semiconductor EDA, Igor Markov, Distinguished Architect at Synopsys
-
Abstract Artificial Intelligence is an avenue to innovation that is touching every industry worldwide. AI has made rapid advances in areas like speech and image recognition, gaming, and even self-driving cars. In the area of chip design, we see a growing gap in what designer’s want to achieve and what the available resources and time allow. This presentation will provide an overview of recent AI trends, show how the latest AI technology including Generative AI can be applied to optimize and automate chip design tasks from architectural design to digital and analog implementation, optimizing silicon life cycles and improving yield. We will also discuss future AI trends and opportunities to accelerate time to market, improve productivity and achieve better quality of results.
- 5. Light Out! Virtualizing the Semiconductor Ecosystem, Joseph Ervin, Managing Director at Lam Research
-
Abstract Artificial Intelligence is a transformative technology that is forcing the semiconductor roadmap to accelerate. Chip complexity is increasing as we move from planar to 3D structures, and smaller and smaller features are making it more difficult to scale and reach yield. How will we produce the next generation of semiconductors? What will the next generation of semiconductor fabs look like?
In this talk, we will review progress toward the “fab of the future”, including where we are now and what is needed to operate a fully “lights out” fab. We’ll discuss some of the requirements of a “smart fab”, including such foundational pillars as autonomous system control, AI-powered smart tools, robotic automation, virtual (digital) twins and virtual device fabrication. We’ll also look at the vast data produced in these new smart fabs, and discuss how artificial intelligence will be a key enabling technology in operating a “lights out” facility. We’ll provide specific examples of how AI and virtual twins are being used to improve chamber productivity and yield, and how humans and AI will work together synergistically in both the physical and virtualized semiconductor fabs of the future. - 6. From Challenge to control: bridging semiconductor process complexity with science-based AI, Changwook Jeong, Associate Professor at UNIST
-
Abstract As semiconductor manufacturing continues to grow in complexity, conventional physics-based modeling approaches—while essential—struggle to meet the demands of multi-physics, multi-scale process integration. In contrast, purely data-driven AI methods offer speed and flexibility but often lack physical grounding, interpretability, and robustness. This talk explores how science-based AI can help bridge this gap by embedding domain knowledge, physics constraints, and uncertainty quantification into machine learning frameworks.
We will focus on three key challenges in semiconductor process modeling: (1) achieving comprehensive coverage across the nine major process modules, (2) capturing multi-scale and multi-physics interactions, and (3) adapting to hidden or evolving process conditions. Through selected examples—including real-time TCAD surrogates, inductive bias for etching, variability-aware modeling, hidden physics discovery, and layout-aware defect and warpage prediction—we demonstrate how science-based AI approaches can enhance both efficiency and physical consistency.
The talk will close with reflections on emerging directions, including the role of physics-aware AI in enabling digital twins and more integrated, adaptive design–process co-optimization.