Just over a year after the launch of its flagship product, Landing AI secured a $ 57 million Series A funding round to continue building tools that make it easier and faster for manufacturers to build and deploy artificial intelligence systems.
The company, started by former Google and Baidu AI guru Andrew Ng, developed LandingLens, a visual inspection tool that uses AI and deep learning to find product errors faster and more accurately.
Ng says that industries should apply a data-centric approach to building AI, which gives manufacturers a more efficient way to teach an AI model what to do, using no code / low code features, which include just a few mouse clicks for to build advanced AI models in less than a day.
“We started the data-centric AI movement and are pretty excited that other companies have started talking about it,” he told TechCrunch. “In manufacturing, each factory does something different, so the problem becomes how we can help 10,000 manufacturers build 10,000 different models without having to hire a lot of labor.”
AI is expected to create $ 13 trillion in realized value for the world economy by 2030, according to a McKinsey study. Ng says not much of it has been realized yet as it remains challenging to build different AI models.
He believes Landing AI has cracked the code to build these models, raising the Series A round after seeing the product market fit and wanting to be able to scale the team to make the product better.
McRock Capital, an investment firm focusing on the industrial Internet of Things, led the investment and was joined by Insight Partners, Taiwan Capital, Canadian Pension Plan Investment Board (CPP Investments), Intel Capital, Samsung Catalyst Fund, Far Eastern Groups DRIVE Catalyst, Walsin Lihwa and AI Fund.
Landing AI has made progress in building its product, but Ng said the company is still in the early stages of the data-centric AI movement and wants to make further progress and innovate on some technology that is still lacking.
For example, while previously building a speech recognition system with 350 million data points, he found that AI technology, invented for many data points, does not work as well in manufacturing settings with limited images to find defects. Part of the data-centric movement is to develop tools to help domain experts use less than 50 images to provide a clearer example of what is a defect.
“We have reached a point where it works and we are eager to scale everything,” Ng said. “We’ve been fascinated for years by how to crack the recipe and take AI to other industries, and we’ve finally gotten there with data-centric AI.”