Plenty of people have a pet project that they are drawn to or consider themselves particularly good at. As the leader of the data science department at Trupanion in Seattle, David Jaw’s projects are actually around pets.
Jaw, GeekWire’s latest Geek of the Week, uses artificial intelligence and machine learning to help automate medical insurance claims for pets, streamlining the process and removing the worry about what’s covered and what’s not.
Born and raised in a suburb near Toronto, Jaw’s family moved to Albuquerque, N.M., when he was 13 years old. He stayed there through college, where he studied mechanical engineering, pursuing a childhood dream of designing airplanes and spaceships.
“Two or three years into undergrad, I realized that my peers could do the aerospace coursework in less than half the time, at a higher quality than I could,” Jaw said. “This relationship was reversed for math and numerical computing classes. I decided to focus on a path that came more naturally to me.”
Jaw moved around for a year or two after college, living off poker winnings.
“Poker started as a solution to a short term problem,” Jaw said. “I was working 30-35 hours per week in a call center for the first half of my undergrad. My grades sucked and my social life was non-existent. When one of my buddies won $15,000 in an online poker tournament, I asked him for guidance, read a few books and quickly turned online poker into a consistent source of income.”
Jaw earned twice as much playing poker as he did in that call center and it remained his “side hustle” throughout his 20s. He even spent a year playing poker full time as an experiment.
“The earnings were higher than a typical engineering job, but not by a large enough margin to offset the downsides of an irregular work schedule and absence of work satisfaction,” he said. “No regrets though. I came away from the experience with greater mental resilience and a group of lifelong friends.”
Grad school and beating people at a card game no longer did it for Jaw and he wanted something new when he moved to Seattle with his now wife. He turned his attention to data science and a job at Trupanion.
“It didn’t take long to notice the opportunity for machine learning to make a meaningful impact at Trupanion,” he said. “We have a strong culture around collecting high-fidelity information and an excellent data warehouse team that executes according to our principles every day. The obstacles we face as a data science team are interesting and satisfying. We answer questions like, ‘How do we replicate our peoples’ thought processes?’ rather than, ‘How do I gain access to the data to begin with?’”
Learn more about this week’s Geek of the Week, David Jaw:
What do you do, and why do you do it? I started and currently lead the data science department at Trupanion.
Medical insurance for pets has relatively low adoption in our country (1 percent) vs Europe and Australia (20 percent+). We believe this is partially due to the clunky claims process. Even if the pet owner has insurance, veterinary bills are paid up front and invoices are submitted for reimbursement. The average processing time is measured in weeks.
In April 2018, we deployed a series of machine learning models that fully emulates the role of a claims adjuster. Each independent thought process that goes into human decision making gets its own code repository and model endpoint. These modular models then feed into an “aggregator” that coalesces all relevant information into a final outcome.
Today, we automate 40 percent of claims submitted through our software at a 99 percent accuracy rate and an average end-to-end run time of 6 seconds. Pet owners can now have their claims settled before they leave the reception counter. We’ve removed the need for policy holders to pay out of pocket, and we’ve removed the uncertainty in not knowing what’s covered and what’s not. Our claims team can now spend their time investigating the complex, interesting situations that utilize their medical knowledge. The easy decisions get automated. No one has lost their job as a result of this shift.
We believe that all pets should get the best veterinary care. No pet owner should have to make treatment decisions with financial burden as a factor.
What’s the single most important thing people should know about your field? Most of us seem to be overestimating the long-term effects of machine learning (murderous sentient AI) and underestimating the short-to-medium-term benefits.
Current AI applications commonly complete their task at surfacing a recommendation. Imagine a time when we have enough confidence in our models to allow physical action to be taken. Rather than seeing a suggested link to a dinner recipe, you could come home to shopped groceries and a meal prepared exactly to your liking. Rather than going to a clinic every year for a flu shot, it could be delivered by a flying needle robot right before an outbreak. Mental resources will be freed up to pursue areas of interest with deeper focus and dedication. Full scale machine learning should be a force multiplier for our civilization.
Where do you find your inspiration? I come up with a lot of ideas on flights. I have a difficult time reading or watching movies without getting motion sickness and my brain kicks into overdrive due a false positive signal of being in danger. Unable to sleep, or consume media, all that’s left to do is take stock of my current position and consider possible paths forward.
What’s the one piece of technology you couldn’t live without, and why? Not the sexiest tech, but my pick would have to be eyeglasses. Life would be far less productive or enjoyable if I couldn’t read a book or screen from further than a few inches away.
What’s your workspace like, and why does it work for you? Our office in south Seattle has hundreds of dogs and cats come in to work every day. It’s difficult to stay in a negative mindset when there is always a cheerful pup or kitten eager for pets a few steps away.
Your best tip or trick for managing everyday work and life. (Help us out, we need it.) I end up thinking about work all the time. I should be the one asking for the advice here, not giving it. Hah!
That said, I’d be even worse off if I didn’t put considerable effort into living a “balanced” life. My approach is to be deliberate in how time and energy is spent. I quantify discretionary time and how it is allocated. I then estimate the effect of time spent in each category/endeavor along with its first derivative. Using this model as a guide, I seek to align time/energy allocation with my personal goals and values.
Mac, Windows or Linux? All of them! Windows for gaming, Mac for general purpose laptop. Linux for deploying ML models.
Kirk, Picard, or Janeway? Can we expand this to any leader in a sci-fi series? If so, I’d pick Chrisjen Avasarala from “The Expanse.” Total badass. I like her willingness to sacrifice personal relationships in service of a greater good. “Earth must come first.”
Transporter, Time Machine or Cloak of Invisibility? Invisibility cloak would be too sneaky to me. Time machine sounds awesome but I would be worried about accidentally causing something horrible, like the extinction of our species. Can’t think of an upside that might balance that out. Transporter wins by process of elimination. The downside is capped at getting stuck or dying.
If someone gave me $1 million to launch a startup, I would … I’d open a noodle shop. I love making food and I would never get tired of iterating toward the perfect bowl.
I once waited in line for … Franklin Barbecue in Austin, Texas. Worth it.
Your role models: My grandmother is the one person I admire most by a large margin. She has been through incredibly difficult times, but never lost her composure. She knew she needed to stay strong for those around her. Despite not having much herself, she would go out of her way to help anyone that needed it. I’m not exaggerating when I say I’ve never seen her do anything selfish or immoral. She is a paragon of what we should all strive to be. Also worth noting is that it is impossible to beat her at mahjong, unless she lets you win out of pity.
She was born and raised in the Sichuan province of China. As the eldest daughter, it was her responsibility to take care of her younger siblings. As such, she wasn’t afforded the luxury of a formal education. She left school at the age of 9 to help with the family farm and prepare meals. When the Communist Party was coming into power, she was in her early 20s with two baby girls. She correctly predicted that things would drastically change in an unfavorable way if she stayed, so she uprooted her entire life and made the arduous journey to start anew in Taiwan. Her sacrifice allowed her two daughters and son to lead happy and fulfilling lives. She repeated this entire process when her children started having their own kids. She again uprooted her life in Taiwan to move to Toronto where she helped raise all of her grandchildren. We now all lead happy and fulfilling lives because of her.
She’s been gone for nine years now, but she will always remain a strong motivation for my decision making. “How can I be better, so that I can live up to the memory of my grandmother?” is a question I often ask myself.
Greatest game in history Poker is my favorite game. Evidence-based decision making is rewarded and cognitive bias is punished.
Best gadget ever: I wear a ring that measures sleep cycles. It should be useful for dialing in an ideal sleep routine, and improving overall health.
First computer: I convinced my parents to buy me a computer for doing homework in high school. I ended up just playing “Starcraft” all day with no remorse. I was a terrible son.
Current phone: Pixel 3.
Favorite app: Audible. Listening to audiobooks makes my long commute enjoyable.
Favorite cause:I like the concept of fairness and I’d support any cause that promotes it.
Most important technology of 2020: The proliferation of machine learning in recent years is due to open sourced tools (tensorflow, pandas, sklearn) and distributed computing as a service (AWS). Without these technologies, it would have been impossible for a small team like mine to get anything impactful done.
Most important technology of 2022: The barrier to entry in utilizing AI is too high. There’s nothing wrong with knowing a lot of math or programming languages, but these difficult-to-obtain skills should not be prerequisites to building good predictive models. Doctors, lawyers or any professional should have the option to enhance their decision making with machine learning. I hope our current position in AI is analogous to the early days of personal computer adoption, where users needed to be familiar with command line execution. Once we hit the GUI operating system phase of the analogy, and the general population gains access to these tools, we’ll start seeing some really interesting applications.
Final words of advice for your fellow geeks: Approach your work with uncompromising integrity. Negative examples like dishonesty and taking credit for others’ work may seem advantageous in the short term, but I’m pretty sure it’s a trap. Humans are exceptional at passively detecting patterns; they will learn to avoid collaboration with bad actors even if they don’t consciously know why. At best, the bad actor successfully takes value out of the system. More likely, the result will be a lower expected value for both the actor and the system. Doing the right thing has the potential to inspire others to do the same, creating a positive feedback loop with unbounded upside.
LinkedIn: David Jaw