We’re spotlighting two on our Information Systems team who are spearheading innovation
Spearheading innovation: That’s been Arrowhead’s brand essence for the past few years. Our mission? To pioneer the most innovative risk solutions to profoundly simplify the insurance experience, using the latest technology.
At the tip of the spear, so to speak, is our Information Systems team, tasked with looking for ways that InsurTech can be used as an Arrowhead differentiator in the marketplace. This diverse team of software engineers, web developers, database developers, business analysts, quality assurance analysts and automation engineers is hard at work, creating easier and more robust ways for producers to do business with us. They’re streamlining both the producer’s and insured’s Arrowhead experience, while at the same time providing us with the data we need to continue evolving.
Today we’re talking with two team members about their roles in Information Systems, how their passion for technology has led them to pursue other projects outside of the office, and how their pursuits have helped create a more dynamic Arrowhead culture.
Tom Lugo, architect consultant for the Batman Experience
A software engineer at Arrowhead since 2015, Tom was asked at a friend’s party to help find programmers to build a project. Turned out the project was to program a virtual reality headset that could be used in a wind tunnel to create the ultimate flying experience. This new friend, Cody Russell, CEO at Tunnel Vision VR had asked dozens of programmers, but all had turned him down.
“Another friend and I told him, sure, we could put together a prototype over the weekend,” Tom recalls.
The prototype was approved, and a few weekends later, the final VR program was completed.
“Now Tunnel Vision VR has eight wind tunnels all over the world: Netherlands, Germany, France, Spain, the UK, Australia, Dubai and Malaysia,” he said. While in the wind tunnel and wearing the VR headset, participants can experience flying over Hawaii or France, for example, Tom explained.
When DC Comics and AT&T brainstormed an ultimate display for the 80th anniversary of DC’s Batman at the 2019 Comic-Con in San Diego, they tasked Media Monks, a production agency, with putting together the virtual reality Batman Experience. Media Monks created the narrative and the visuals, calling on Tunnel Vision VR to help with the programming.
“It was electric,” Tom recalls. “The whole weekend the Batman Experience was at maximum capacity, even though we were located a couple of miles from the main Comic-Con event.”
Forbes said, “The main attraction at The Batman Experience is the VR skydiving booth called ‘The Dark Knight Dive,’ a virtual reality skydiving interactive experience allowing you to fly through Gotham City, where you’ll encounter many of Batman’s villains and allies.”
The New York Times called the wind tunnel “The centerpiece: Dark Knight Dive, a skydiving attraction with a virtual reality element.”
Wired Magazine said, “Comic-Con’s Batman VR experience soars – while others fall flat.”
What’s next with Tunnel Vision?
Tunnel Vision plans to take the show on the road to multiple U.S. cities, he said. They’re also working on a virtual reality/augmented reality project with HALO jumps (high altitude, low opening) for the military.
“In training, these guys must work through 18 life-threatening situations, where something goes wrong. It’s much easier to train using a wind tunnel for simulation than for each person to jump 18 times out of a plane to experience and solve the 18 training situations.”
What Arrowhead projects are you working on?
“I’m really excited about a lot of current projects. Right now we’re creating this ‘magic’ API layer that will put us at the forefront of how producers will be writing business in the future. We’re focusing on aggregators and web services to start building out these microservices, making our systems more flexible and reactive. In another project, we’re revamping and reorganizing the policyholder side of Arrowhead Exchange, making it mobile friendly, more dynamic, more flexible. And in a third project, we’re working with a third-party vendor to create a more robust system that streamlines the process to discover the appetite for various programs without having to manually look up SIC codes.”
Arrowhead will soon be creating its first hackathon, he added, saying they’re currently in planning stages.
“In a nutshell, I’m a software engineer that works on the Strategic Initiatives Team,” he said. Tom joined Arrowhead four years ago just out of the Art Institute of California, where he received his BS in Advertising. He served three combat tours in Iraq, Bahrain and Afghanistan as an Intelligence Analyst and Psychological Operations Sergeant in the US Army. Currently Tom’s working on his MBA while occasionally taking on weekend jobs as a solutions architect consultant.
Jason Antic, creator of DeOldify
“As a web developer, I’ve wanted to get involved in “deep learning” (artificial intelligence) for awhile,” Jason explained in a recent interview. “It’s a technology that’s become very popular. A lot of companies are pouring money into similar projects.”
Jason proved that just one person and his gaming computer can create a deep learning project: He created DeOldify, an algorithm that will automatically colorize and restore old black-and-white photos and videos.
“I took a course at Fast.ai – a free course and one of the best ones I’ve ever taken. My capstone project at the end of the class led to DeOldify.”
Up to this point, automatic colorization was primitive at best, or had to be done manually.
How DeOldify was hatched – and went viral
“It took me six weeks to get an initial working model last October (2018). When it finally worked, it was state-of-the-art. I was so excited about the results that I posted about it, including the source code for free, on Reddit. It was picked up by BoingBoing.net who wrote an article – and then the project went viral.”
Jason told his wife, Lyndi, who also works for Arrowhead as a senior SQA (software quality assurance) analyst, “Oh no, this is going viral, and I’m going to have to give talks, and life will be very different!”
Sure enough, Jason has been invited to speak at two Mensa events – one in San Diego and also at their national conference. He also spoke at this year’s F8, Facebook’s Developer Conference. He’s been a guest speaker on podcasts and has given several interviews.
“I released the code for free, so anyone can grab it and use it. And DeOldify is one of the most popular AI projects online right now – right up there with Google’s and Facebook’s AI projects.”
Jason and his Fast.ai instructor, Jeremy Howard, have completed a second project, inventing a new, more robust version of generative adversarial networks – GANs.
“Traditionally, a deep learning model just consists of a model that teaches itself to complete certain tasks. But a GAN includes two models: the critic which teaches a student, and the student which tries to fool the teacher/critic. They both gradually learn in the process,” he explained.
Prior to their work, GANs started with two models that basically knew nothing and produced nonsense. But Jason and Jeremy started by teaching each model separately using straightforward methods, “teaching” them as far as they could go. Only afterwards, they put them together to interact with each other with normal GAN training.
“Traditionally, GANs are very volatile and can easily go off the rails by teaching each other garbage. This is because they’re starting from scratch- they both know nothing at the beginning of training.” Jason explained. “So we created a way to keep them from teaching each other said garbage, by giving them an education first. Basically, we start by teaching them independent of each other as far as we can take that, before we have them teaching each other. Otherwise it would be like the blind leading the blind. Or perhaps it’s like expecting a community of feral children to reinvent the world’s science and culture in a generation. That’s not going to happen!”
Can the DeOldify algorithm be applied at Arrowhead?
“There’s a lot more that can be done with deep learning– not just colorizing! Deep learning is a very general approach with wide-sweeping applications: images, sound, text, database, natural language processing and a lot more. One example project that I’ve mused about at Arrowhead could clarify and simplify project management: Train a deep learning model so that you could give a text description of a project and it could tell you how long a project will take. It can remove bias and present a truer picture of the project without having to involve people doing estimates, saving a lot of time. The same exact approach could also be used to help underwriters estimate risk on a quote.”
“I received my bachelor’s in computer science at Penn State. I was in the National Guard in college and worked in Iraq for awhile as a weather forecaster.”
Fresh out of the military, Jason and his new wife Lyndi moved to San Diego. He started at Arrowhead almost immediately (12 years ago) in software quality assurance, then went on to software engineering and web development.
Arrowhead is becoming synonymous with spearheading innovation
Historically, Arrowhead has focused on innovation, says Tom Kussurelis, chief marketing officer. The company is currently involved in InsurTech and its various applications in the insurance industry, looking for ways to offer more insurance programs and simplify the insurance procurement process.