Skip to content
Leo Zhang
Open to opportunities

LeoZhang

Software engineer focused on building products that scale.

I've spent more than ten years shipping production software across mobile, enterprise platforms, IoT devices, and AI systems—leading teams, integrating hardware, and delivering solutions from concept to production.

Post-Study Work Visa — Open full-time work for any employer, valid until 05/2029
Leo Zhang

About Me

I started my career writing C++ streaming cores and ASP.NET admin backends for a small IoT start-up in China. Over the next decade, the systems grew with the company — from live-streaming chat rooms to factory automation, from enterprise messaging to AI research.

Every project taught me the same lesson: the hard part isn't writing code — it's making sure it keeps working when the network drops, the database outgrows the server, or the only person who understands it leaves. I've designed for those failures across 21 projects in 6 domains — factory MES, enterprise IM, IoT gateways, mobile field apps, live streaming, and applied AI.

I recently completed a Master's at AUT in Auckland with First Class Honours, building a multimodal AI virtual try-on system and publishing at IVCNZ 2025. Now I'm looking for a team that values systems that actually work under real constraints — not just on a whiteboard.

0

projects shipped — factory MES, enterprise IM, IoT, mobile GIS, AI research

0st

Class Honours — Master's at AUT, thesis completed 6 months early, published at IVCNZ 2025

0+

factory sites — smart manufacturing platform serving hundreds of workers daily

0

daily active users — enterprise messaging, <200ms latency, <2% downtime over 10 years

0%

defect elimination — legacy C++ IM to cloud migration, zero-downtime cutover

What I Build

Capability areas shaped by a decade of production delivery

Mobile Systems

10 years of Android — from NDK streaming cores to offline field apps. I build for the device, not the demo.

Kotlin
Java
NDK/JNI
Compose
MVVM
UART/GPIO

AI & Vision

Deployable AI — models that run on real hardware, not just notebooks. Diffusion on Raspberry Pi, classification at the edge, LLM agents with Dify.

PyTorch
YOLO
LoRA
Diffusion
DeepSeek
OpenCV

Backend & APIs

Enterprise backends for factory automation and messaging — Spring Cloud, event-driven design, MySQL at millions of rows. Built for uptime, not resumes.

Spring Cloud
MySQL
Redis
Kafka
ActiveMQ
REST

IoT & Hardware

Systems that bridge software and physical devices — scales, locks, conveyors, gateways. Serial protocols, GPIO, RS485. Environments where you can't push a hotfix.

MQTT
RS232/485
RFID
GPIO
Modbus
Raspberry Pi

DevOps & Delivery

CI/CD pipelines that keep a small team shipping. Docker, Jenkins, GitLab — boring tools that work. Cloud migration and repatriation when the math says so.

Docker
Jenkins
GitLab
Nginx
Linux
CI/CD

Selected Work

Each project taught me something about building systems under real constraints

ChatClothes Virtual Try-On

ChatClothes Virtual Try-On

Master's thesis — a multimodal virtual try-on system combining lightweight YOLO classification with LoRA fine-tuned diffusion models and LLM agent control. Built on limited resources: two 16GB Tencent Cloud free-tier GPU servers (one for diffusion, one for LLM) + Raspberry Pi 5. Published at IVCNZ 2025. 94.2% accuracy, FID 28.5, completed 6 months early.
PythonPyTorchOOTDiffusionLoRA
FID 28.5 (19%↑), 75% hand artifact reduction, 94.2% accuracy, 87% user success (50 users)
Smart Factory Platform

Smart Factory Platform

Garment factories needed real-time production tracking across multiple sites. 6-person team. I owned all Android clients (phone/tablet/shop-floor), wrote ~1/4 of backend (3 Spring Cloud modules), built hardware communication programs (scales via serial-to-WebSocket bridge, conveyors, washers), and helped with Vue.js frontend when needed. 10+ factory sites, 30%+ efficiency gain, hundreds of daily workers.
JavaSpring CloudMyBatisAndroid
10+ factory sites, 30%+ efficiency gain, 6-person team, hundreds of daily workers
Enterprise Messaging Platform

Enterprise Messaging Platform

Internal messaging tool for 5,000 employees had chronic stability issues. Legacy C++ core was unmaintainable — migration had to happen without downtime or message loss. Incremental migration to Spring Cloud + Kotlin Android client, keeping the NDK transport layer for low-latency delivery while replacing everything around it. 90%+ defect elimination, <200ms latency, 500K+ daily messages, ran stable for 10 years.
KotlinJavaC/C++NDK
5,000 DAU, <200ms latency, 500K+ msgs/day, 90%+ defect reduction, <2% downtime over 10 years