HK image pushes homegrown AI startup to move to US

Ben Ng March 10, 2020 12:38

Tech entrepreneurs Taylor Host and Jamie Wilde, who have resided in Hong Kong for years, founded artificial intelligence startup Miro AI in 2017, using AI and computer vision technology to analyze bodies in motion and provide insights for athletes and sports brands.

In two years since its launch, the firm has gained traction in business, with clients from Europe, North America and Asia. It decided to relocate to the United States after realizing that being a Hong Kong entity is turning off some Western investors.

“We have always been proud to be a Hong Kong company, but the current brand no longer works for us,” Host, Miro AI’s chief executive, told EJ Insight in an interview.

Miro AI helps amateur and professional runners find their running event photos instantly with its RunnerTag product, and gathers data for sportswear companies to enable them to target consumers more effectively.

Backed by Silicon Valley-based SOSV, Miro AI started in Hong Kong because of its geographical location, as well as the broad base of sports enthusiasts in the city and the large-scale marathon and trail running events being held here. 

“One of the requirements for good quality image analysis with deep learning is a large training set – a lot of data. Running was a place where we could get these huge volumes of data – it's a consistent, repeatable data source. For us, finding that data streams was critical,” said Wilde, the company's chief technology officer.

“There are other companies finding their data streams from different places, maybe from security data, but we found a high-quality data stream in running.”

With most of the technical development having been conducted in Hong Kong, Miro AI’s RunnerTag is the fastest, most-accurate, bib reading, athlete identification technology available, according to the firm.

Event organizers upload their event photos to the platform and in seconds, the service identifies the participants in each photo by bib number.

Used in more than 750 events around the globe to power their event photo services, RunnerTag has identified roughly 8 million athletes with the service, “and we can use this data, both in pseudonymized and anonymized forms, to fuel our AI model development”, Wilde said. 

All this data contributes to Miro AI’s body data set. “We have taken in this raw, pseudonymised data of athletes, and we have structured it very tightly using proprietary AI and traditional computer vision,” Host said.

“Pairing the fresh visual data provided by events which use our AI with volume, we are the only company in the world that is able to recognize running shoe brands on people's feet as they are running by.”

35 million shoe samples

Miro AI has now over 35 million shoe samples in its dataset to train its AI models, and one of the applications is tracking the runners’ shoes and outfits for event organizers, sponsors, and sports brands.

"We are making analyses on the ground predictions about who is wearing these shoes – by performance, pace, and basic demographics,” said Wilde, “tracking the progress of new market entrants in running shoes… Just using sales data, a brand like Nike cannot understand the effect of the Chinese new market entrants.”

Host said the firm now has around 30 to 35 sports brands using its AI-powered services, including Nike, Adidas, and Asics. “Within the top 10 running brands in the US, we’ll count half as clients by year-end.”

The vast majority of Miro AI’s business is in the United States and Europe, while it also has a growing number of clients in Japan, South Korea, and Southeast Asia, according to Host.

The firm has crossed the US$1 million mark in gross revenue, and expects a bigger revenue this year.

“We have chosen a very narrow vertical within sports to begin building – running. We saw an opportunity within sports to immediately begin generating revenue to offset our costs,” said Taylor.

“We were bootstrapped for nearly two years. We have since raised a small amount of money, but we are a team of ten that is largely supported by revenue right now. That has made us a lot less susceptible to these economic and political issues.”

Despite that, months of increasingly violent protests and a bruising US-China trade and tech war have brought Hong Kong’s economy to a recession for the first time in a decade last year, while pushing Miro AI to move its headquarters to the United States.

Need for capital

“Coronavirus, the Hong Kong protests, as for any growing company, are challenges,” said Taylor. “Our decision to move our headquarters to the US was driven primarily to position the company better for raising venture capital. We currently have investments from the US, Australia, and Hong Kong, so it’s not only important for us to have sizeable US presence, but also to be a US company.”

Miro AI set up a new holding entity in Delaware late last year, but its engineering, product development, and operational entity remain in Hong Kong, according to Taylor.

Miro AI has also established a partnership with New York-based photography website provider Photoshelter, as the firm will be working with visual data from baseball games and teams in Major League Baseball (MLB).

For the upcoming MLB season, Miro AI and Photoshelter have built AI models to power real-time image analyses of baseball games for multiple teams in the league.

“It’s an exciting time and we’re geeked at the opportunity to build AI for the highest levels of sports,” said Taylor.

“Geopolitics play a role in business; startups are not insulated. Unfortunately, the political attitude globally is exclusive: you are either pursuing Chinese investments or you are pursuing non-Chinese investments. Raising capital from China can exclude you from certain investors elsewhere in the world,” he said.

“We have to make the choice, and so far, with the high concentration of our business in non-China markets, the choice has been clear.”

– Contact us at [email protected]

CG

Ben Ng

EJ Insight writer