PengyaoChen

For the last 15 years, I've grown consumer products to hundreds of millions of people at Instagram, Facebook and Uber by personalizing them with machine learning.

LLMs take personalization a step further.

Now at Commure, I lead engineering for Ambient AI that listens to a clinic visit and writes the note in the doctor's own voice, built on LLM orchestration and evals.

Let's talk See my badges ↓

The badge drawer

Commure

2026–

Engineering Leader, Ambient AI

Lead the full-stack Ambient AI org of 15 backend, frontend and ML engineers, owning the Ambient AI Scribe end to end.
Orchestrate LLM calls that turn a visit recording into clinical-grade notes: high recall, low hallucination.
Built an eval harness that pairs code checks with LLM judges, grounded in real clinician feedback.
Capture on iOS, Android, web and a Chrome extension, with robust audio chunking and self-healing uploads; clinicians tune each note to their own style.

Uber

2019–26

Staff Software Engineer, Membership (2019–21)
Engineering Leader, Member Lifecycle (2021–26)

Bootstrapped Uber’s membership, from Ride Pass and Eats Pass to Uber One, and grew it to 50M members and about half of Uber’s gross bookings.
Led the Member Lifecycle org, 20+ backend, ML, mobile and web engineers across Retention and Experience, driving paid-to-paid retention above 90%, best in food delivery.
Built incentive levers, a configuration platform and a neural-network targeting model that doubled incremental gross bookings per dollar versus generic promotions.
Championed AI adoption with tech talks and a hackathon: 100% of engineers coding with AI and 10+ custom tools built.

University of Michigan

2010–11

B.S.E., Computer Engineering

Capstone: a four-way superscalar, out-of-order Alpha R10K-style CPU, synthesizable at 80MHz, with fast branch recovery, a speculative load/store queue, instruction prefetching, a victim cache, a BTB and a Gshare predictor.
Built an automated regression suite in shell and a visual debugger in C; passed all 29 test cases with the lowest simulated CPI of any team, 1.3.
Interned on Skype’s partner integration team, porting SkypeKit to embedded Android and Linux.