I’m Cory Moore. I build AI systems for hiring, and I test whether they work.
My training is in industrial-organizational (IO) psychology, the field that studies people at work. The part I care about most is measurement: how you assign a score to something (or someone) on some attribute, and show that it means what you claim it means. I finished my PhD at North Carolina State University in 2025, where I studied psychometrics. Through the PhD I also worked as a data scientist in hiring tech, applying what I was learning to real products.
Now I’m an Applied Scientist at Hirevue, which makes AI software for hiring and assessment. I design and prototype LLM systems, such as agents that score interview answers against a rubric. I also build the evaluations that decide whether those systems are ready to ship. Before Hirevue, I spent two years running psychometric analyses on high-stakes tests, including the military’s entrance exam, at the Human Resources Research Organization.
Most people who build AI systems don’t come from measurement, and most measurement people don’t build AI systems. I work between the two. Psychometrics spent a century learning how to tell whether a human rater can be trusted. Many of those methods apply almost unchanged to evaluating AI systems and developing LLMs that grade another model’s output. I think AI evaluation has a lot to learn from psychometrics, and most of what it needs is already worked out.
This site is where I work through those ideas: evaluation, LLM judges, agent design, and AI in hiring. The views here are mine, not my employer’s.
Outside work, I’m in the mountains as often as I can be. I live in Colorado, where I trail run, ride gravel, and backpack. My longest race so far is a 50-mile trail race with 11,000 feet of climbing. This season I’m training for a 100-mile gravel race, and a half Ironman is on the list for 2028. I track my training the way I’d track a model: I build training plans and a pipeline that pulls my workout data and checks whether the plan is actually working.