AI, to Be Precise: Brooklyn Company Detects Manufacturing Anomalies with Tools Powered by Deep Learning

Nanotronics uses Quadro RTX to accelerate its deep learning optical inspection platform up to 15x faster than CPU.
by Nicole Castro

Nanotronics is pushing the limits of precision manufacturing with AI powered by NVIDIA Quadro RTX technology.

The Brooklyn-based company combines super-resolution images with deep learning in its nSpec line of advanced optical inspection tools.

Delivering a wide range of scanning methods and imaging techniques, these tools help customers find anomalies, defects and other features of interest throughout production workflows.

Nanotronics’ platform combines AI, automation and 3D imaging so customers in industries like semiconductors and materials and those operating research and development facilities can capture and analyze anomalies in real time and in much higher quality.

Nanotronics’ nSpec product finding a silicon carbide defect.

With NVIDIA Quadro RTX 8000 and 6000 GPUs, Nanotronics has accelerated its platform by up to 15x compared to CPUs, while retaining accuracy and performance. NVIDIA RTX technology also freed up the CPU for other computations.

The architecture of the RTX 6000 and 8000 allows for the efficient use of different processing levels and faster analysis, enabling the nSpec tools to be agile and customizable without sacrificing any processing capabilities.

“Nanotronics adapts its models to the continuous improvements of NVIDIA architectures with new core types,” said Matthew Putman, co-founder and CEO of Nanotronics.

NVIDIA Quadro RTX Improves Accuracy, Flexibility and Performance

Nanotronics’ platform uses computer vision and deep learning to streamline inspection processes. The AI captures defects and anomalies with more accuracy than human vision.

With these deep learning models, the company can achieve a high level of accuracy in near real-time processing. But Nanotronics wanted to integrate more complex solutions to help customers meet increasing demands in their workflows, such as scanning a larger number of samples at much faster rates.

NVIDIA Quadro RTX provides more memory and processing cores, which enable the nSpec products to easily integrate into production pipelines to scan, analyze and report results for objects in bulk, like components with large surface areas and volumetric form factors.

For example, when running deep learning analysis for anomaly and defect detection on semiconductor wafer scans, Nanotronics saw the Quadro RTX 6000 performed 15x faster than conventional enterprise CPUs.

A sample report showing defect maps of a wafer.

“The development of deep learning algorithms using GPUs has provided solutions for some of the most crucial technological challenges faced by our customers,” said Putman. “With NVIDIA GPUs, our detection models function 12 to 15 times faster than any previously available methods.”

Using NVIDIA GPUs, Nanotronics can deploy an enhanced AI solution for detection in industrial workflows, allowing customers to increase the efficacy of quality assurance for high-precision manufactured goods.

Nanotronics will be at the American Aerospace & Defense Summit on Dec. 11-12 in Glendale, Ariz., to discuss the latest developments of intelligent manual manufacturing.