HomeAutomobileAndrew Ng Startup, Touchdown AI, Speeds Manufacturing unit Inspection

Andrew Ng Startup, Touchdown AI, Speeds Manufacturing unit Inspection



Pc imaginative and prescient specialist Touchdown AI has a novel calling card: Its co-founder and CEO is a tech rock star.

At Google Mind, Andrew Ng turned well-known for displaying how deep studying may acknowledge cats in a sea of pictures with uncanny pace and accuracy. Later, he based Coursera, the place his machine studying programs have attracted practically 5 million college students.

As we speak, Ng is finest recognized for his views on data-centric AI — that bettering AI efficiency now requires extra give attention to datasets and fewer on refining neural community fashions. It’s a philosophy coded into Touchdown AI’s flagship product, LandingLens.

Based in 2017, Touchdown AI counts amongst its customers Foxconn, StanleyBlack&Decker and automotive provider Denso. They and others have utilized deep studying to enhance their effectivity and cut back prices.

A Classification Problem

A chip maker with manufacturing vegetation across the globe was one of many first to strive LandingLens. It needed to make use of deep studying to enhance throughput and yield of the wafers that carry chips by its fabs.

Like all chip makers, “they’ve lots of visible inspection machines on the fab ground that scan wafers at varied steps — and so they do a superb job discovering anomalies — however they didn’t do as properly classifying the issues they discovered into forms of defects,” mentioned Quinn Killough, Touchdown’s liaison to the client.

And like many chip makers, it had tried a wide range of software program applications for classification. “However the options wanted to be fine-tuned for every product and with greater than 100 merchandise, the funding wasn’t value it,” mentioned Killough, who has a background in laptop imaginative and prescient and manufacturing.

AI Automates Inspection

Then the client utilized AI with LandingLens. It’s designed to deal with the end-to-end MLOps course of — from accumulating information to coaching and deploying fashions — then handle the continuing means of refining the fashions, and particularly the information, to reinforce outcomes.

Though it’s nonetheless early days for the deployment, the product and its data-centric strategy have already helped the chip maker cut back prices.

“The first engineer driving the challenge mentioned he sees deep studying as transformative and desires to scale it out throughout his facility and get different vegetation to undertake it,” mentioned Killough.

Inspectors within the Cloud

The chip maker used LandingLens on NVIDIA V100 Tensor Core GPUs in a cloud-based service that runs inference on a whole bunch of hundreds of pictures a day.

“We weren’t certain of the throughput capabilities at first, however now it’s clear it could actually deal with that and much more,” mentioned Killough.

The identical service can prepare a brand new classification mannequin in lower than a minute utilizing about 50 defect pictures so customers can iterate quickly.

“On the coaching aspect, it’s essential for our instrument to really feel snappy so our clients can troubleshoot issues and experiment with options,” he mentioned.

Taking AI to the Edge

Now the corporate is taking the AI work to the manufacturing facility ground with a brand new product, LandingEdge, which is in beta exams with a number of clients.

It captures pictures from cameras, then runs inference on industrial PCs geared up with NVIDIA Jetson AGX Xavier modules. Insights from that work feed on to controllers that function robotic arms, conveyor belts and different manufacturing programs.

“We intention to enhance quality control, making a flywheel impact for quick and iterative AI processes,” mentioned Jason Chan, product supervisor for LandingEdge.

Accelerating a Startup’s Development

To get early entry to the most recent know-how and experience, Touchdown AI joined the NVIDIA Metropolis program, geared for corporations utilizing AI imaginative and prescient to make areas and operations safer and extra environment friendly.

It’s nonetheless early days for the corporate and data-centric AI, which Ng believes could also be one of many greatest tech shifts on this decade.

To be taught extra, watch a GTC session (free with registration) the place Ng describes the standing and outlook for the data-centric AI motion.

RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Most Popular

Recent Comments