Explorer · AI-Native Advocate · Behavior Designer
I build and evaluate AI products, connecting quality, cost, and user value, and advocate for the workflows and culture that help teams use them well.
DropboxSlackTikTokUber
01 · Bio
Leon builds and evaluates AI products, and designs the systems that help organizations adopt them well. His work draws on behavior design, data science leadership, causal inference, and machine learning.
At Dropbox, his recent work centers on AI product evaluation: defining model and product quality, connecting it to user outcomes, and measuring cost and latency alongside product value. He also designs reusable workflows, shapes adoption strategy, and rigorously validates decision models.
02 · Selected work
Selected themes from recent work across workflow design, product measurement, and decision science, generalized here for public sharing.
Prototyped a governed approach for making individual AI workflows reusable, with clear ownership, human review, attribution, and adoption feedback loops.
Developed a measurement framework spanning quality, latency, cost, engagement, and experimentation, brought together through dashboards and an interactive metric tree for product decisions.
Pressure-tested a business-driver model by separating robust signals from weak associations, tiering the evidence, and defining causal follow-up around what was known, what remained uncertain, and what to test next.
03 · Writing
Practical notes on building AI products, measuring impact, and making models useful in real systems.
Browse all writing04 · Contact
Reach out to talk about AI-native workflows, product measurement, behavior design, data science leadership, or causal inference.
leonzhao705@gmail.com© 2026 Leon Zhao