I design AI-native enterprise systems where people remain decision makers. I turn ambiguous AI and technical strategy into product direction: clear workflows, prototypes, and decision-making artifacts that help teams decide what to build, before they invest.
I've spent my career helping people make sense of complex systems: understanding what is happening, knowing what to trust, and staying in control when things go wrong. That challenge is now central to designing AI teammates.
AWS IoT · 2024–2025 · North Star UX Vision Co-Lead
Customers spent a year or more stitching together 18+ separate services. I initiated and co-led a North Star vision: an intelligent layer that turns a customer's goal into a working solution, so the job becomes describing an outcome, not assembling the pieces by hand.
AWS IoT · 2023–2024 · Design Lead · 1:20 designer-to-engineer
A new lightweight runtime for constrained devices, launched without fragmenting the platform. Sole designer across three engineering teams. I simplified device setup and pushed safety validation in without slipping the launch.
Amazon Web Services · 2026–present · Design Lead
The hard part of an AI teammate is not what it can do. It is whether a person can undo it when it is wrong. A study with 21 people found trust depends on reversibility, not capability, and I used that to frame when the AI acts on its own, when it asks first, and when a person steps back in.
Overview →AWS IoT · 2020 · Design Lead · 50+ engineers
Led end-to-end UX for a complete platform re-architecture, shipped on schedule as the design lead supporting 50+ engineers. I restructured the backend's tangled data model into three concepts users could act on. With no console spec, I derived UX requirements directly from the API architecture, so engineering never waited on design. Shipped with zero frontend-backend mismatches. Demoed by James Gosling (Creator of Java) at re:Invent 2020.
Before the current Generative AI wave: in 2016, I shipped a conversational AI shopping assistant powered by IBM Watson for companies like The North Face. Overview →
I turn unclear AI opportunities into concrete workflows and prototypes that teams can evaluate and decide on.
I make technical systems understandable, trustworthy, and something people can act on.
I align product, engineering, and leadership around a shared direction, then leave behind frameworks they can use without me.
I built a personal AI assistant trained on my own notes and how I work, and I use it every day. It teaches me how its memory works, how to correct it, how trust builds, and how working with AI changes over time.
My toddler found the AI's wrong answers hilarious, not frustrating. It was a small reminder that users and makers rarely experience the same product.
Read the write-up ↗︎Taught myself hardware programming by building a tic-tac-toe game, experimenting with three development tools and AI code generation. The gaps I hit gave me firsthand empathy for customers wrestling with edge device constraints.
Read the write-up ↗︎Led a three-person team in an AI hackathon to prototype automated customer support for Greengrass. This experiment later informed my advocacy for intelligent orchestration in the IoT Experience Vision.
Internal work.
Every action an AI teammate takes spends from a trust budget: a small one for a suggestion, a much bigger one for taking your cursor. I ran a study simulating 21 people working with an AI teammate, and the designer's job is to help that trust build up, a little at a time.
Read the post ↗︎All 12 said the same thing: give intent, not adjectives. Edit like a director, not a client. And never hand over the decision.
Read the post ↗︎How I treat AI as a teammate rather than a tool, and five practical tips from my day-to-day work, from giving context to keeping AI honest.
Read the post ↗︎Read more essays and build logs at blog.yingyingz.com ↗︎.