Hi, I'm Yunpeng.
I build AI agents that turn “who knows how this works?” into a one-minute answer.
I'm a Staff Software Engineer at Apple. I build AI agents, and the harnesses that make them dependable, so engineering teams can understand their services, infrastructure, and processes in minutes, just by asking in plain language.
Ideas → shipped- years building infrastructure and developer platforms
- 13+
- for an engineer to get answers about an unfamiliar service, just by asking
- minutes
- Apple News automation pipeline run time
- hours → minutes
- developer throughput from PR-based concurrent testing
- 2×
the short version
About me
I'm happiest in the gap between an idea and its first working prototype. This site is my workshop: what I'm building, what I'm learning, and the systems I'm making simpler.
Right now I build AI agents for engineering teams at Apple, along with the harness around them: the tools, context, and guardrails that make an agent dependable. An engineer asks a question in plain language and gets a grounded answer about their services, infrastructure, internal knowledge, and processes in minutes, instead of spending days reading wikis and tracking down the one person who knows.
Alongside that I build AI-powered productivity tools that shorten the whole path from idea to release: faster development, faster qualification, and fewer manual steps between a commit and production.
All of it sits on more than a decade of platform engineering. I lead DevOps and infrastructure work for Apple News, Weather, and Stocks. Before that I built next-generation CI infrastructure at Uber's Advanced Technologies Group, and earlier at Apple I took the News automation pipeline from hours to minutes.
I think like a founder and prototype like one. I can take a rough idea to something people can use quickly, and I care about design as much as architecture, because a tool only matters if people enjoy using it.
A few facts
- This site runs on a six-node Raspberry Pi Kubernetes cluster at home, deployed with Argo CD.
- Most of my projects start as a weekend prototype. The good ones don't stay prototypes.
- I started my career in test automation, which is why I still care how things fail.
- Engineering Systems & Computing graduate of the University of Guelph, Canada.
- The assistant on this site is one of my own experiments in retrieval and agentic workflows.
How I work
Start with a prototype
A working demo settles more debates than a design doc.
Simplify complex systems
The best platform is the one nobody has to think about.
Design is part of engineering
If a tool isn't a pleasure to use, it won't get used.
AI where it earns its place
Agents grounded in real systems, not demos.
Own it end to end
From the first sketch to production monitoring, the whole path is the product.
right now
What I'm focused on
Building AI agents and agent harnesses at Apple, so engineers can learn a service, its infrastructure, and its processes in minutes by asking in plain language.
Building AI-powered productivity tools that speed up the development and release cycle, from first commit to production.
Prototyping new ideas on nights and weekends, this site included: its assistant, the Kubernetes cluster it runs on at home, and whatever I get curious about next.
Updated March 24, 2026
so far
The journey
- Feb 2020 – PresentStaff Software Engineer · Apple Inc.
- Feb 2020 – 2024Senior Software Engineer · Uber Advanced Technologies Group
- Mar 2016 – Feb 2020Senior Software Engineer · Apple Inc.
- Jun 2015 – Mar 2016Software Development Engineer · IBM
- May 2014 – Jun 2015Software Developer · IBM
- Oct 2013 – May 2014Sr. Automation QA Engineer · Tangelo Games Corp. (Imperus)
go on, ask
Ask anything about my work
Ask about infrastructure, projects, writing, current work, or experience.
off the clock
Snapshots

I used this display picture for quite some time.

I miss you a lot.
made to order
Resume, your way
don't be a stranger

