Senior Android Developer · Java Backend · Full-stack Engineer
📧 leozhang2056@gmail.com | 📍 portfolio https://portfolio.leoz.fun 🔗 LinkedIn: linkedin.com/in/leo-zhang-305626280

Senior Software Engineer with 10+ years of experience and Master's in Computer and Information Sciences from AUT, graduating February 2026. Delivering enterprise solutions across Android development, Java backend systems, IoT platforms, and AI/ML integration.

Proven track record evolving a standalone IM client into a multi-subsystem enterprise platform over 10 years, leading cross-functional teams up to 20 people, deploying IoT solutions managing 1000+ devices, and architecting digital signage platforms for large-scale retail networks.

Key Strengths:

• Android Expertise: 10+ years building enterprise Android apps — from IM clients (NDK TCP/UDP) to IoT gateways (serial protocol, MQTT) to digital signage players, with hardware integration (RFID, scales, cameras, serial devices)
• Platform Architecture: Evolved a messaging tool into a full enterprise platform (user center, file service, permissions) serving 5,000 DAU over 10 years — from sole developer to leading 20-person team
• IoT & Edge Computing: Designed Android-based gateways bridging non-networked industrial equipment (circuit breakers, smart meters) to cloud platforms via MQTT, managing 1000+ devices with offline resilience
• Backend Development: Spring Cloud microservices, time-series databases (InfluxDB), stream processing (Apache Flink), distributed file storage (FastDFS)
• AI-Assisted Development: Daily use of Claude Code, Cursor, and GitHub Copilot for code generation, refactoring, and review — comfortable integrating AI tools into the full development workflow
• Additional Skills: ML model training (PyTorch, LoRA, YOLO), BI platforms (PowerBI, FineBI, Kettle ETL), hardware protocol integration (Modbus, RS485, UART), .NET desktop development

Master's Student & AI Researcher | Jul 2024 – Feb 2026

• ChatClothes — Developed diffusion-based multimodal virtual try-on system with YOLO classification and LLM-driven interaction
  • Optimized for edge deployment on Raspberry Pi 5 with full offline capability
  • Integrated ComfyUI + Dify for workflow orchestration
  • Thesis passed with First Class Honours; completed ahead of schedule

Technical Lead / Senior Software Engineer | Jan 2013 – Jun 2024

Started as sole developer, grew to lead 20-person cross-functional team. Owned Android development, architecture design, and platform evolution across IoT, manufacturing, and enterprise communication projects.

Key Projects:

• Enterprise Messaging Platform — Sole developer → team lead over 10 years (2014–2024)

  • Started as the only developer, built first Android IM client and pushed to production
  • Designed custom TCP protocol (WebSocket was unstable), NDK transport for sub-200ms latency
  • Evolved standalone IM into multi-subsystem enterprise platform (user center, file service, attendance, approvals) serving 5,000 DAU
  • Migrated from unstable C++ messaging to Easemob cloud, eliminating 90%+ defects
  • Still in use and actively growing

• Smart Factory Backend — Full-stack tech lead (2018–2024)

  • Led 6-person team delivering Spring Cloud microservices across 5+ factory sites
  • Built hardware communication (electronic scales via serial-to-WebSocket bridge, conveyors, RFID)
  • Deployed Android shop-floor terminals and Vue.js dashboards for hundreds of daily workers
  • Improved production efficiency by 30%+; awarded Hebei Provincial Science & Technology Award

• Smart Broadcast Control Platform — PM & Android Team Lead (2020)

  • Led 9-person team, delivered 1000+ device Digital Signage platform in 2 months ahead of schedule
  • MQTT broadcast for large-scale ad delivery to retail displays (bookstores, clothing stores)
  • Managed high-concurrency MQTT connections — primary technical challenge at scale

• IoT Solutions — Sole Android developer for 7-year product line (2016–2023)

  • Smart switches and gateways: Android app + embedded gateway firmware + Spring Cloud backend
  • Participated in architecture design with time-series database and message queue

• Smart Power Management — Java backend & Android developer (2019–2022)

  • 20-person company-wide project: 3+ parks, ~15% energy reduction, sub-second alarms
  • Integrated Modbus/RS485 gateways and smart meters with InfluxDB + Apache Flink

Started as sole developer, grew to 20-person team. Custom TCP protocol (WebSocket was unstable), NDK transport for sub-200ms latency. Evolved from standalone IM to multi-subsystem enterprise platform.

Key Features:

• Custom TCP/UDP protocol with NDK transport achieving <200ms latency
• Migrated from unstable C++ messaging to Easemob cloud (90%+ defect elimination)
• Platform evolution: IM → user center → file service → attendance, approvals, asset management
• 5,000 DAU, 500K+ daily messages, <2% downtime over 10 years

Tech Stack: Kotlin, Java, NDK, Custom TCP/UDP, Spring Cloud, Easemob, FastDFS, MySQL, Redis
Impact: Still in use and actively growing after 10 years

Microservice-based platform deployed across 5+ factory sites, connecting brands, factories, and workers. Won Hebei Provincial Science & Technology Award.

Tech Stack: Spring Cloud, Vue.js, Docker, Jenkins, MySQL, Redis, RFID, Electronic Scales
Impact: 30%+ efficiency improvement, 99.9% uptime, hundreds of workers supported daily

Digital Signage platform for retail stores — led 9-person team, delivered in 2 months ahead of schedule. MQTT broadcast for large-scale ad delivery.

Key Features:

• 1000+ Android display devices across bookstores, clothing stores, design shops
• MQTT broadcast for high-concurrency ad campaign distribution
• Three-component architecture: Vue.js Admin Console, Customer Portal, Android Device Client
• Device fleet management with heartbeat monitoring and offline cache/recovery

Tech Stack: Android, MQTT, Spring Boot, Vue.js, ExoPlayer, SQLite
Impact: Replaced manual on-site content updates, delivered ahead of schedule

7-year product line: smart switches and gateways sold to customers. Sole Android developer for all mobile apps. Participated in architecture design with time-series DB and message queue.

Tech Stack: Android, Embedded Linux, Zigbee, WiFi, MQTT, Spring Cloud, InfluxDB, Message Queue
Impact: End-to-end IoT product from gateway firmware to cloud platform to mobile app

20-person company-wide project: enterprise power monitoring for factories, buildings, and campuses with real-time data collection and energy analysis.

Tech Stack: Spring Cloud, InfluxDB, Modbus, RS485, MQTT, Apache Flink, Vue.js, ECharts
Impact: ~15% energy reduction, <1s alerts, 99.9% data collection rate, 3+ parks deployed

Master's thesis project at AUT. Developed diffusion-based multimodal virtual try-on system with YOLO classification and LLM-driven interaction. Published at IVCNZ 2025.

Tech Stack: PyTorch, Diffusion Models, YOLO12n-LC, DeepSeek LLM, ComfyUI, Dify, Raspberry Pi 5
Impact: Published at IVCNZ 2025, 19% FID improvement over baseline, deployed on edge devices

General-purpose AI data annotation and model training platform, validated with 20-class herbal recognition demo. Users upload data, publish annotation tasks, train models, and deploy inference.

Tech Stack: Python, FastAPI, YOLO, ResNet, MobileNetV3, Docker, GPU Training
Impact: End-to-end platform workflow; launched but not maintained post-launch

Predictive maintenance platform analyzing historical equipment records to predict maintenance windows using Random Forest and time-series analysis.

Tech Stack: Python, SparkNet, Random Forest, Time-series Forecasting
Impact: Shifted maintenance strategy from fixed-cycle/reactive to predictive

Led performance optimization across 8+ enterprise Android apps over 8 years. Systematic optimization across 6 domains: memory, CPU, UI fluency, cold-start, APK size, stability.

Key Achievements:

• APK 200MB+ → 80-90MB (55% reduction) via R8/ProGuard and SO stripping
• Cold-start ~1.5s → ~800ms (47% improvement)
• Systematic OOM/ANR governance across 8+ apps

Tech Stack: Kotlin, Java, C++, LeakCanary, Android Profiler, R8, ProGuard, Heap Dump
Impact: Measurable performance improvements across entire app portfolio

In-house hotfix framework (AndFix → Tinker) for emergency patch deployment. After years of practice, led the decision to retire it in favor of staged rollout + feature flags.

Key Decision:

• Identified unsustainable maintenance cost (patch-per-version model, expanding QA matrix)
• Retired after proving staged rollout provided better ROI
• Evolution: AndFix → Tinker → Full Upgrade + Staged Rollout

Tech Stack: Kotlin, Java, Method Replacement, Dex Merge, Canary Deployment, Feature Flags
Impact: Technical evaluation and retirement decision saved long-term maintenance overhead

Smart gateway controlling non-networked circuit breakers via RS232 serial bus. Each gateway manages up to 70 breakers. Connects to Smart Power Management Platform.

Tech Stack: Android, RS232, Modbus RTU, MQTT, Edge Alerting, InfluxDB
Impact: 1:70 gateway-to-breaker ratio, 24/7 operation, multi-site deployment

School face recognition attendance using ArcSoft SDK for on-device comparison. 3-person team — I built the Android terminal. 1-2 seconds per student, 99%+ accuracy.

Tech Stack: Android, ArcSoft Face SDK, Liveness Detection, Gate Control, JPush
Impact: Deployed across multiple schools, real-time parent notifications

24/7 self-service borrow/return system for schools with face recognition, QR auth, electromagnetic locks, and UV disinfection. 10+ hardware peripherals via UART/RS485/GPIO.

Tech Stack: Android, Face SDK, ARM Controller, UART/RS485, GPIO, Spring Cloud
Impact: <3s average borrow time, deployed in multiple schools, zero book loss

Mobile patrol system for forest rangers in no-signal areas. Custom offline maps, GPS tracking, GIS risk-point annotation, and deferred sync on network recovery.

Tech Stack: Android, GDAL, Geos, Shapely, GPS, Offline Maps, SQLite
Impact: Enabled reliable patrol data collection in zero-connectivity forest areas

Service robot for exhibitions — bought robot with SDK, built features on top: navigation, face recognition, voice interaction, knowledge base. 4-person team.

Tech Stack: Android, ArcSoft Face SDK, iFlytek ASR/TTS, Spring Cloud, SLAM
Impact: Deployed at client exhibition hall, still in use

Hardware data acquisition (serial ports), ETL (Kettle), and BI (FineBI) for a banknote paper production facility. Fully air-gapped environment.

Tech Stack: SQL, Pentaho Kettle, FineBI, Serial Port, UHF Scanning, Linux
Impact: Data pipeline from scattered databases and hardware devices to unified BI layer

Early career multi-language live streaming platform (C#, C++, Node.js, Python, Lua) with RTMP/HLS/WebRTC.

Tech Stack: .NET, C++, Node.js, Python, Lua, RTMP, WebRTC, FFmpeg
Impact: 1,000+ peak concurrent viewers, 99.5% streaming uptime

Early-career solo project: enterprise patent document management with Lucene.NET full-text search, SQL Server, and Excel batch processing.

Tech Stack: C#, .NET Framework 2.0, Lucene.NET, SQL Server, Windows Forms
Impact: 10,000+ patents managed, <2s search response, 50+ daily users

Android cryptocurrency trading app with real-time market data via WebSocket, secure wallet storage, and exchange/recharge flows.

Tech Stack: Kotlin, Jetpack Compose, Room, Retrofit, WebSocket, MVVM
Impact: Real-time market data, 3-layer security architecture

C# application for highway toll station vehicle weighing, overload alarms, and invoicing system integration.

Tech Stack: C#, Electronic Scale Integration, Alarm System
Impact: Deployed at highway toll station

Multi-terminal visit booking and access management for high-security scenarios (prison, hospital). WebRTC video visitation with server-side time enforcement.

Tech Stack: Spring Cloud, WebRTC, Face Recognition, WeChat Mini Program
Impact: Deployed across multiple departments, supported pandemic contactless visits

Shared technologies and domain relationships across all projects:

graph LR
  subgraph Enterprise
    IM["Enterprise Messaging\n(10-yr platform)"]
    SF["Smart Factory\n(5+ sites)"]
    SP["Smart Power\n(3+ parks)"]
    BC["Broadcast Control\n(1000+ devices)"]
  end

  subgraph IoT
    VG["Visual Gateway\n(1:70 ratio)"]
    IoTD["IoT Solutions\n(7-yr product)"]
    PBL["Picture Book Locker\n(10+ peripherals)"]
  end

  subgraph AI
    CC["ChatClothes\n(Diffusion + LLM)"]
    CH["Herbal Recognition\n(YOLO/ResNet)"]
    DM["Maintenance Prediction\n(Random Forest)"]
  end

  subgraph Mobile
    Boobit["Boobit\n(Compose)"]
    SA["School Attendance\n(Face SDK)"]
    FP["Forest Patrol\n(Offline GIS)"]
    ER["Exhibition Robot\n(SLAM)"]
  end

  subgraph Govt
    BP["Banknote Paper Mill\n(Air-gapped ETL)"]
    VS["Visit System\n(WebRTC)"]
  end

  %% Shared technology links
  IM -- "Spring Cloud\nKotlin/Java" --> SF
  SF -- "MQTT\nVue.js" --> BC
  SP -- "MQTT\nInfluxDB" --> IoTD
  IoTD -- "RS485\nModbus" --> VG
  VG -- "MQTT\nEdge alerts" --> SP
  CC -- "YOLO" --> CH
  CH -- "Python" --> DM
  SA -- "Face SDK" --> ER
  PBL -- "Face SDK\nUART" --> SA
  FP -- "Android\nOffline-first" --> IoTD
  BP -- "C#/.NET" --> VS
  BC -- "Android\nFleet mgmt" --> IoTD

┌─────────────────────────────────────────────────────────┐
│                     CAREER BY NUMBERS                   │
├──────────────┬──────────────┬───────────────────────────┤
│  10+ yrs     │   5,000 DAU  │   1,000+ IoT devices      │
│  Experience  │  Messaging   │   managed concurrently     │
├──────────────┼──────────────┼───────────────────────────┤
│  20-person   │   5+ factory │   ~15% energy reduction    │
│  team led    │   sites      │   across 3+ parks          │
├──────────────┼──────────────┼───────────────────────────┤
│  <200ms      │   99%+ face  │   10,000+ patents          │
│  latency     │   recognition│   indexed                  │
│  (TCP/NDK)   │   accuracy   │                            │
├──────────────┼──────────────┼───────────────────────────┤
│  55% smaller │  47% faster  │  99.5% streaming uptime    │
│  APKs        │  cold start  │  (Live Commerce)           │
└──────────────┴──────────────┴───────────────────────────┘

Master of Computer and Information Sciences
Auckland University of Technology (AUT), New Zealand | 2024 – Feb 2026

• First Class Honours · GPA 7.41/9.0
• Research Focus: AI, Computer Vision, Edge Computing
• Thesis: ChatClothes Virtual Try-On System — passed ahead of schedule

Bachelor of Software Engineering
Hebei University of Science and Technology, China | 2009 – 2013

• National Scholarship recipient

• Zhang, Y. et al. ChatClothes: Conversational Virtual Try-On with Diffusion Models — IVCNZ 2025 · DOI 10.1109/IVCNZ67716.2025.11281834
• Zhang, Y. Clothes Recognition Based on Lightweight Deep Learning Models — Aesthetics in Creative Technology, IGI Global · 2026
• Zhang, Y. ChatClothes: A Virtual Try-On System Using Conversational AI and Diffusion Models — AUT Master's Thesis, 2025 · AUT Repository

Last updated: June 2026