Host & fabric
Dedicated silicon, high-speed networking, and storage I/O boundaries wired directly to cloud fabric—zero hypervisor layer.
Architecting cloud and networked systems from bare metal up—with an eye for clarity, craft, and tactile interaction.
Software Engineer on Google’s GCE Bare Metal team and UCLA Ph.D. in Computer Science. Every interactive unit below is a domain-accurate micro-simulation of real infrastructure and research work.
01 / CURRENT FOCUS · GOOGLE CLOUD INFRASTRUCTURE
On Google’s GCE Bare Metal team, I focus on how dedicated physical machines behave as first-class citizens of a cloud platform—bridging hardware boundaries, automated fleet lifecycle, and predictable fabric performance across the stack.
Dedicated silicon, high-speed networking, and storage I/O boundaries wired directly to cloud fabric—zero hypervisor layer.
Automated provisioning, firmware/health verification, live orchestration, safe draining, and reclaim across physical fleets.
Predictable low-jitter packet flows, fault isolation, and operability at scale from host NIC through control plane.
Ultra-low-latency trading platform
One place this architecture comes together is the low-latency trading platform built through the Google Cloud and CME Group collaboration, which I contribute to on the GCE Bare Metal side.
The collaboration brings dedicated bare metal compute and deterministic networking to financial markets—combining the precision of physical infrastructure with the flexibility of the cloud.
Platform context from public Google Cloud and CME Group materials.
02 / INTERACTIVE PLAYGROUND · 7 DOMAIN MICRO-APPS
Every project card is a custom domain instrument: fracture a 5G link for SEED, probe an eSIM policy barrier, trace an AST call graph, decode NB-IoT waterfall bursts, reprogram Flora’s 4-layer stack, test wireless VR motion-to-photon lag, and map 2,208 Acemap topics live.
Diagnosing and recovering from 5G failures through the SIM.
Authentication and fine-grained access control for SIM and eSIM.
Explore a codebase through questions, context, and GPT.
# [1] entry.py -> [2] auth.py -> [3] context.py
token = parse_bearer(req.headers)
claims = auth.verify(token)
return context.bind(claims, req.repo)
Starts at entry.py, verifies bearer credentials
in auth.py, and binds the caller scope inside
context.py.
An open NB-IoT network for experiments with real devices.
A software-defined mobile network spanning eSIM, radio, core, and edge.
Proactive LTE uplink scheduling for interactive mobile VR—less delay, more immersion.
An academic knowledge graph mapping connections across 2,208 research topics and 19 disciplines.
03 / PUBLISHED RESEARCH · 2019–2023
No publications match that filter. Clear the search box or switch back to All.
04 / A LITTLE ABOUT ME
I care as much about how a system behaves under the hood as how an interface feels in someone’s hands.
Before Google, I received my Ph.D. in Computer Science from the University of California, Los Angeles, advised by Prof. Songwu Lu, and my B.S. in Electronic Engineering from Shanghai Jiao Tong University. My research spans networked systems, mobile security, and edge and cloud infrastructure—tracing how real systems fail, recover, and perform.
Alongside systems engineering, I’m drawn to product interaction and visual aesthetics: clear hierarchy, purposeful motion, and interfaces that make complex technical ideas feel effortless.