arXiv preprint · September 2026
SSD-LLaMA: SSD-Native Inference for Trillion-Parameter MoE at 1+ Token/s on a Consumer PC
Coordinating SSD, RAM, and VRAM for large-model inference on personal hardware.
ZILI MENG / RESEARCHER & BUILDER
HONG KONG · EXPLORING WHAT’S NEXT ↗AI, madepersonal.
Personal AI powered by nearby compute, on the devices you already use.
Explore WiCiChoose a floating idea to explore.
A few places along the way.
SELECT A LOGO TO EXPLOREHKUSTAssistant Professor · 2023–present
Leading SPARK Lab, working on AI systems and real-time intelligence.
Full journey01 / BUILDING
Where most of my energy goes.
WiCiPersonal AI.
Personal AI.
Untethered.
I’m building WiCi, the wireless compute layer for personal AI. The idea is simple: give the devices we already use access to powerful, nearby GPU compute.
My focus spans LLM inference, GPU–memory–storage systems, and the wireless stack that connects them.
Meet WiCi Read our research ↗
WiCi One
02 / RESEARCH
All publicationsIdeas are a start.
Make them work.
How can AI run beyond the datacenter?
How can networks respond at the pace of people?
These questions connect my research.
8 selected works
↗
arXiv preprint · August 2026
WiCi: Wireless GPU Computing Infrastructure
Bringing nearby GPU compute to the devices people already use.
arXiv preprint · revised May 2026
Make a Video Call with LLM: A Measurement Campaign over Six Mainstream Apps
CallArena: measuring quality, latency, and overhead in real-time AI video conversations.
ACM SIGCOMM
Interpreting Deep Learning-Based Networking Systems
Understanding the decisions learned by neural systems.
ACM SIGCOMM
Learning Scheduling Algorithms for Data Processing Clusters
Learning to make better scheduling decisions for data-processing systems.
arXiv preprint · April 2026
OCC: Physical-Layer Assisted Congestion Control for Real-Time Communications
Using physical-layer information to make interactive communication more responsive.
USENIX NSDI
Law: Towards Consistent Low Latency in 802.11 Home Networks
Making the networks closest to us more responsive.
ACM SIGCOMM · IRTF ANRP 2026
ACE: Sending Burstiness Control for High-Quality Real-time Communication
Better delivery for interactive experiences.
ZILI MENG 孟子立HONG KONG ↔ THE WORLD
03 / THE PERSON BEHIND THE WORK
A little curiosity.
A lot of building.
I’m an Assistant Professor in Electronic and Computer Engineering at HKUST, with a courtesy appointment in Computer Science and Engineering. I lead SPARK Lab.
I received my B.Eng. and Ph.D. from Tsinghua University. Along the way, I’ve been fortunate to learn from people at MIT, Carnegie Mellon, and across industry. Today, that background in learning, networks, and systems comes together in my work on AI.
I care about the whole path from a research idea to something people can actually use.
04 / PEOPLE & POSSIBILITIES
Teaching & service ↗Made of ideas.
Built by people.
At SPARK Lab, we work across AI, systems, and networks. We’re looking for curious researchers and engineers who want to turn ambitious ideas into working systems.
Recognition,
shared.
Across my research and the students I mentor.
Research impact
IRTF Applied Networking Research Prize
2026 · ACE
Best Paper Awards
IWQoS 2021 · ICC 2020
Coauthored research
SIGCOMM gold
Student SRC Gold · 2024 & 2025
Hyun Hu Park & Samuel Yeung
SRC Gold · 2018
My student research
Doctoral work
Doctoral dissertation awards
ACM China · CIE · Tsinghua
ByteDance Scholar · 2022
Fellowships & scholarships
Microsoft Research Asia · 2020
Ph.D. Fellowship
Tsinghua Top Grade Scholarship · 2018
HAVE AN IDEA?
Have a good
what if?
AI systems, research collaborations, or building something with WiCi—I’d love to hear from you.
zilim@ust.hkRoom 2441, Academic Building
HKUST, Clear Water Bay, Hong Kong