SemiAI, Co., Ltd.

Dr. Tae Kwon Jee, CEO , SemiAI, Co., Ltd.

Taekwon Jee is the Founder and CEO of SemiAI, an AI startup building agentic AI solutions for semiconductor manufacturing. SemiAI’s SMILE platform integrates prediction, root-cause analysis, engineering copilots, and physics-informed simulation to help semiconductor fabs improve yield, accelerate process optimization, and reduce engineering workload. 

Before founding SemiAI, Taekwon held semiconductor engineering and technology roles at SK hynix, ASML, Lam Research, and Samsung, gaining extensive experience across semiconductor processes, equipment, and advanced manufacturing.

He earned his Ph.D. from the University of California, Berkeley. Under his leadership, SemiAI has launched multiple proof-of-concept projects with global semiconductor manufacturers and equipment companies, demonstrating strong performance in virtual metrology, overlay prediction, defect analysis, and AI-based diagnostics.

SemiAI is a member of NVIDIA Inception and was selected as a finalist in the 2026 Lam Capital Venture Competition. Taekwon’s vision is to build an AI operating system that enables semiconductor engineers and autonomous AI agents to work together in smarter, faster, and more scalable fabs.

 

Topic:

Agentic AI for Predictive Yield and Autonomous Fab Analytics

 

Abstract:

This presentation introduces a multi-peer agent architecture for predictive yield improvement and autonomous fab analytics. Instead of relying on a single centralized AI, specialized agents independently analyze process, equipment, metrology, defect, and historical data, then exchange 

evidence and validate one another’s hypotheses. Through continuous collaboration, the agents predict yield risk at wafer, lot, chamber, and recipe levels, identify likely root causes, simulate corrective actions, and recommend the highest-impact interventions.

 

The approach combines virtual metrology, anomaly detection, causal reasoning, and physics-informed simulation in a closed-loop workflow. By moving from isolated prediction to coordinated diagnosis and action, fabs can achieve higher prediction accuracy, faster root-cause analysis, reduced engineering workload, and more consistent yield improvement. The result is a scalable foundation for autonomous semiconductor manufacturing, where multiple AI agents work as an integrated engineering team.

 

Key Technologies Covered:

• Multi-Peer Agent Architecture 

• Predictive Yield Modeling 

• Virtual Metrology 

• Wafer-Level Risk Prediction 

• Autonomous Root-Cause Analysis 

• Process and Equipment Anomaly Detection 

• Cross-Domain Fab Data Integration 

• Physics-Informed Synthetic Data Generation 

• Hypothesis Generation and Multi-Agent Validation 

• Closed-Loop Yield Optimization and Action Recommendation