Semiconductor

A*STAR I2R’s Semiconductor Division develops AI-native digital technologies to accelerate semiconductor design, manufacturing, and production. We focus on resource-efficient, data-driven AI solutions that address high-mix, high-complexity semiconductor challenges. Working closely with foundries, equipment vendors, and national initiatives, we translate advanced research into deployable tools that improve yield, shorten development cycles, and enable next-generation semiconductor innovation.

Our focus areas include: 

AI-Driven 2D/3D Metrology & Inspection

We develop data-efficient AI for high-resolution 2D and 3D metrology and inspection, enabling faster defect detection, classification, and root-cause analysis. Our solutions address challenges in advanced high-mix manufacturing, delivering robust performance with limited labels and enabling scalable deployment in both lab and fab environments.

Intelligent Design of Experiments & Process Optimisation

Our AI-powered DoE and optimisation engines accelerate process development by identifying optimal recipes with fewer wafer runs. By combining sparse-data learning with physics-informed models, we enable faster learning cycles, improved yield, and first-pass success across complex semiconductor processes such as etching, CMP, and advanced integration.

Advanced Process Control & Predictive Manufacturing

We build AI-based process control and monitoring solutions that move beyond reactive rule-based systems. By integrating multi-sensor data, virtual metrology, and anomaly detection, our approaches enable predictive fault detection, reduced false alarms, and adaptive control for high-mix, high-variability semiconductor manufacturing environments.

TinyAI & Edge-Optimised AI Systems

We develop hardware-aware AI compression and optimisation technologies for ultra-efficient edge deployment. Our TinyAI solution enable low-latency, low-power inference on constrained devices by co-optimising models and hardware, supporting scalable deployment across industrial, embedded, and next-generation edge-AI applications.