Ishan Jha

Technology today is at a crossroads between Industry 4.0 and 5.0, where products and services are manufactured and designed to be ‘human-centered’ and sustainable. Such objectives have always been expected in education, especially in the way the workforce is trained. Leveraging on prior research which was done by a team from Pasadena City College (PCC) and UC Irvine, we have formed a bi-coastal collaborative, including Mercer County Community College (MCCC) and Princeton University, to expand and transcend our earlier AI-powered virtual reality (VR) simulation framework and platform, AI-powered digital twin (DT/AI) for education. This Phase-2 effort demonstrates a working, multi-institution R&D platform with the following capabilities: (a) new training modules with enhanced lithography training and advanced device packaging, (b) better optimized, high-fidelity equipment emulations and process simulations that much closely replicate the physical equipment, (c) adherence to documented facility-specific Standard Operation Procedure (SOP) and manufacturing process flow, (d) a parallel fabricated chip-under-test and its test board for I/O connectivity verification, and last but not the least, (e) a tightly-coupled agentic AI engines available throughout the training session to customize learner-centric experiences to enhance individual knowledge acquisition and retention. The authors also continue to demonstrate the feasibility and scalability benefits of foundational VR/AR-based training for students, technicians, and up-skill learners for whom direct cleanroom or packaging lab access is often unattainable; while with DT/AI, learning and feedback is affordable and widely accessible.

VR simulations hold great potential for training, and AI capabilities can scaffold the learning process through large language model (LLM) powered tutors that provide personalized responses to students. However, such training typically requires multidisciplinary teams of programmers, 3D artists, and instructional designers, and complex software such as game engines. If such advanced training could be created by less skilled users using simpler software, then a vast amount of quality training would become available to help address the workforce challenges. Here, we investigate whether it is feasible for community college and K-12 students to create such VR simulations. We describe how the students in the author group designed and created an AI-powered VR simulation for a common semiconductor fabrication process and pilot-tested it to evaluate its effectiveness. Pilot test results show that students reported high levels of engagement, ease of use, and increased interest in and knowledge of semiconductor manufacturing. Linear regression analysis of the data revealed that engagement, ease of use, knowledge increase, and interest increase are significant predictors of the desire to use VR training in the future. These findings show that given well-designed software tools and expert mentoring, a team of community college and K-12 students can successfully produce engaging and effective VR training simulations. Our work suggests a nationally scalable approach to addressing the semiconductor workforce challenge while also training the next generation of VR simulation developers. To our knowledge, this is the first time K-12 and community college students have developed and evaluated VR simulations with high efficiency and engagement.