How do we build AI agents that earn user trust, execute tasks autonomously, and defer to human judgment when the stakes are high?

My career started at Snowbird Ski & Summer Resort, doing snow safety work and directing operations as the Chief Aerial Tram Operator. Twenty years later, the same question that drew me to mountain operations is why I show up to work today: how do people, technology, and organizations intersect when the stakes are real?

At Snowbird it was my responsibility to ensure that 5,000+ daily guests had the experience of a lifetime while ensuring safety as a top priority. In the mornings, when it had snowed overnight, we would transit to the top of the mountain in the Aerial Tram scouting out pockets of snowpack instability. Then we would strategically ski the mountain throwing hand explosives and doing ski cuts setting off smaller avalanches to safely navigate to Howitzer guns. We used the Howitzer guns to mitigate avalanche risk in areas of high risk and high consequence safely from a distance. On those mornings before sunrise, I'd witness how the right outcome depends on humans, technology, and multiple organizations working in tandem when there is significant risk.

While working full-time at Snowbird, I was also pursuing my Honors Bachelor of Science in Business Information Systems at the University of Utah. I was drawn to how technology helps people and organizations work in harmony. After graduation, I joined Wasatch Global Investors as a Systems Administrator. There I kept our on-premises and remote data centers, workstations, and mobile devices fully operational to power multi-billion dollar portfolios. Downtime had real consequences. A missed trade meant a hit to someone's retirement account and a ransomware attack would have made the news for the wrong reasons. Similar to Snowbird, the work was about coordination across people, systems, and departments. When any single piece failed, the consequences impacted people who trusted us.

After some time in IT, I no longer wanted to just work at the intersection of humans, technology, and organizations. I wanted to study this domain and figure out how to make improvements. At the same time, my physician parents had planted a deep-rooted interest in medicine that I couldn't shake. This inspired me to obtain my EMT certification to be an effective first responder while working at Snowbird. On the side, I also helped at my dad's pain management clinic doing IT work.

That's how I learned about the field of health informatics, which studies how to improve human health through technology. Early in my PhD program I encountered UX research, and in that moment I knew I'd found my calling. For my dissertation I designed, built, and evaluated a conversational agent, called Hernia Coach, for post-surgical patient education. A couple of years later, that work became the foundation for a CHI paper I co-authored with academic and industry researchers, extending Nielsen's heuristics for conversational agents.

When I graduated with my PhD I subsequently joined Microsoft Azure. The work looked different on the surface. Azure customers were running mission critical cloud infrastructure their businesses depend upon, not patients learning about hernias. But the same fundamental research questions remained. How can you build appropriate trust in a system? When should users intervene, and when should the system act autonomously? How do you mitigate the risk of high impact consequences? These are questions I'm still exploring, especially in my agentic AI studies.

Great research generates the need for more research. Let's keep digging in.