Learning runs deep with AWS DeepRacer event

Machine learning is a subset of artificial intelligence that enables computers to learn through experience. 

Securian Financial already leverages machine learning in a variety of ways. For example, machine learning models may be embedded in the technology tools and platforms we purchase, while other models may be developed and trained by Securian. Ultimately, machine learning will enable higher levels of automation and increase our operational efficiencies.

“Artificial intelligence and machine learning will have many different applications at Securian Financial, including understanding handwritten text and fraud detection,” says Tammy Quant, senior director of the Technology Innovation Lab. “We also envision using machine learning to create predictive models to improve our customers’ experiences.”

The Technology Innovation Lab is where Securian Financial tests emerging technologies and builds new capabilities concepts for digital businesses. This includes technology like blockchain, artificial intelligence, machine learning and other cutting-edge technology. The Lab brought AWS DeepRacer to Securian to help prepare for future uses cases involving machine learning.

What’s AWS DeepRacer?

AWS DeepRacer is an autonomous model-sized race car. It takes machine learning one step further by testing reinforcement learning models (a machine learning technique) on a 26-foot race track.

Teams of developers and data scientists “train” the car in a cloud-based simulated environment by reinforcing desired behavior. Cameras on the car view the track, and a reinforcement model controls the throttle and steering.

“The majority of employees involved in DeepRacer had little to no prior experience working directly with machine and reinforcement learning,” says Brian King, who works in the Technology Innovation Lab and helped bring AWS DeepRacer to Securian Financial. “A team from AWS provided some great educational content on the first day that helped explain those concepts.

Brandon Burggraff is an engineering consultant at Securian Financial who participated in the event.

“It was a lot of test and learn,” he explains. “Each member of my team came up with their own function, and then we all ran simulations to find which model we wanted to refine. We then let our models train overnight [in the virtual environment] and loaded them onto physical cars the following morning. After a couple of trial runs, we chose our final model for the actual races, and we ended up in first place!”

Employee participation and key takeaways

Around 20 employees were selected to participate in the event.

“This event served as validation that sometimes simple models are most effective,” says Julia Rodd, a data science consultant at Securian Financial and DeepRacer participant. “It’s natural to try to make models more complex by throwing the kitchen sink at them. But for this event, we had limited time to train our models. We knew we needed to focus our time in certain places and not build anything that would need additional training. Like anything, it’s important to have a strategy and to stick to it.”

Brandon Burggraff had a similar realization.

“The most interesting takeaway for me was how easy it was to over-train a model, and how the biggest key to success is to let your car learn to optimize itself,” he says. “It’s almost a direct contrast to building robust code that can handle any situation. It’s better to code something to walk and hope it learns to run, compared to trying to code it to run from the get-go.”

The opportunity to learn new skills in a unique environment provided many benefits for participants.

“It’s important to keep having events like DeepRacer to give employees the opportunity to gain new skills, meet other employees and leverage existing skills in different ways,” Julia says. “It’s nice to get away from the desk for a little bit to practice those skills, too — these forums create an environment where employees can bring these new skills back to their current roles.”

“Having events like DeepRacer on campus are great ways to bring hands-on experiences in new technologies for our employees,” Brandon says. “It demonstrates Securian Financial’s commitment to investing in its people and promotes innovation from within.”

Needless to say, the event was a success

“This event was a great way to engage our employees in cutting-edge technology and get exposure to complicated processes like machine learning and reinforcement learning,” says Brian Oberman, director of engineering, Technology Innovation Lab. “One of the major takeaways was for employees to start thinking about how they could apply what they learned from DeepRacer to our business needs here at Securian Financial.”