An inside look at how research firm Morningstar is racing machine learning-powered model cars to help solve complex data problems

Morningstar

Research firm Morningstar is using Amazon Web Service’s DeepRacer, a machine learning-powered model race car, as a training tool. That comes as Wall Street firms have been talking more about the public cloud and artificial intelligence.
James Rhodes, Morningstar’s chief technology officer, told Business Insider that by using the cars, employees have in turn developed new ways to help analysts extract data from tables in financial statements.
Chicago-based Morningstar created a company-wide league, and 450 employees have signed up.
Morningstar now has permanent race tracks at four of its locations, and plans to send its company champion to compete at the AWS re:Invent conference in December.
Click here for more BI Prime stories.

Pop into Morningstar’s Chicago headquarters and you might see something unexpected: A group of employees cheering on what appears to be a remote-controlled toy car racing around a track.

It’s not all fun and games, though. Morningstar, one of the world’s biggest investment research companies, is turning to machine learning-guided cars to learn about ways to better pull and analyze data.

Amazon Web Services has been using the DeepRacer cars to introduce clients on its public cloud services to machine learning technology. Wall Street firms, meanwhile, are talking more about wading into the public cloud and uses for artificial intelligence.

James Rhodes, Morningstar’s chief technology officer, said he was introduced to the AWS DeepRacer when employees asked if they could spend their training stipends on the cars.

The model race cars use so-called reinforcement learning, an advanced form of machine learning that relies on trial-and-error instead of pre-set training data. Rhodes said that by training with the cars, his team has in turn developed ways to help analysts automatically extract information from tables in financial statements.

"You get one, and it’s like, ‘Ok, this might be a personal interest of this person,’" Rhodes told Business Insider. "But when you start getting four, five, six, you start looking into it a little bit more."

"I realized that it was a really effective tool that we could leverage to give our staff certain skills that we’re looking for," he said.

Rhodes said the cars offer experiences in three important areas: they give employees exposure to machine-learning techniques, and people get to test out programming languages they might not see in their day-to-day jobs, such as Python, which is commonly used in data science. They also get hands-on experience with the public cloud.

"It’s like the trifecta of training," Rhodes said. "I get to expose people to newer techniques. I get to expose people to programming languages I want, and I get to expose people to the infrastructure we’re moving towards. And I get to do it in a fun way."

‘All in’ with tracks, racing leagues

AWS started a racing league in late 2018, which culminates in a championship at re:Invent, the cloud provider’s conference in December. More than 150 customers are set to participate, with 10% from the financial services industry, an AWS spokesperson said.

Morningstar "went all in" on the AWS DeepRacer, Rhodes said. It created an internal racing league for the cars, with a promise that the overall champion would get flown to Las Vegas to compete in re:Invent.

Despite offering no monetary incentives to participate, 450 employees signed up, forming almost 100 teams in groups of three to six that spanned 10 countries.

A majority came from Morningstar’s tech team — about 35% have signed up — but people across the business are involved. The research team has also shown heavy interest, Rhodes said.

"It’s the new way of people interfacing with data," said Rhodes of the general interest in machine learning. "The old way would be everybody dumped everything in Excel."

"Now, I’m seeing much more of a trend in non-technical groups of people leveraging things like Jupiter notebooks and Python," he added.

Users can test and train AWS DeepRacers in a virtual environment, but Morningstar has pulled out all the stops with physical setups.

Four offices — Chicago, London, Mumbai and Shenzhen — have dedicated spaces for permanent physical tracks, which measure 26′ by 17′. Morningstar’s other six locations have temporary tracks.

"Those are some very interesting initial conversations with the facilities teams," Rhodes said about discussing office space needs for the race tracks.

See more: Wall Street is finally willing to go to Amazon’s, Google’s, or Microsoft’s cloud, but nobody can agree on the best way to do it: ‘If you pick a favorite and you’re wrong, you’re fired’

Employees quickly learned that there was a risk of stray cars shooting down hallways as people fine-tune their models. The world record for completing a single lap around the 59-foot course is 7.44 seconds, which is 5.4 MPH or about 100 MPH when scaled up to a real-size car.

Portable barriers were added to prevent accidents, Rhodes said.

There are team logos and names, including "The Reinforcer," a nod towards reinforcement learning, "Need for Speed" and even "Can’t Drive."

In Chicago, the company is also hosting another league that includes three other large financial firms. Rhodes declined to name who else was involved, but said the league will start in the coming weeks, and the plan is to create similar ones at other Morningstar locations.

The internal league kicks off the first week of October, and winners in all 10 Morningstar locations will be crowned. In November, the company will invite the 10 regional winners to Chicago for the Morningstar championship.

"It’s very competitive. There’s a fair amount of trash talking," Rhodes said.

"But in a friendly way," he quickly added.

NOW WATCH: Violent video games are played all over the world, but mass shootings are a uniquely American problem