Wan Tsz Kin (Dr. Wan Tsz Kin), Senior R&D Software Engineer

Dr. Wan Tsz Kin

Wan Tsz Kin — LLM serving & production AI platforms

Senior R&D at Huawei · vLLM-Omni core contributor · Hospital Authority healthcare AI · PhD, City University of Hong Kong

Get in touch GitHub LinkedIn

About

Shipping models to production

Wan Tsz Kin (also known as Tsz Kin Wan) focuses on distributed inference for LLM and multimodal models—coordination, load balancing, and benchmarking—with end-to-end healthcare AI delivery experience in Hong Kong.

Current

Huawei AI Platform & Framework · Senior R&D Software Engineer (2024–)

Research

Biomedical / health AI research affiliation with CityU; peer-reviewed publications; stroke assessment patent co-inventor

Languages

Cantonese (native) · Mandarin (fluent) · English (professional)

Research

Publications & patent

Peer-reviewed work in biomedical AI, clinical prediction, and health data science — plus a machine-learning stroke assessment patent.

Patent

  • Co-inventor · 2021

    Method and system for machine learning-based assessment of stroke

    Chan, K. H. K., Huang, R., Liu, J., Wan, T. K., Woo, Y. M. P.

Selected publications

  1. Life · 2022 · First author

    Identifying predictors of COVID-19 mortality using machine learning

    Wan, T. K. et al. · Life, 12(4), 547 · DNN, SVM, RF, XGBoost

  2. Computers in Biology and Medicine · 2023

    Stroke mortality prediction based on ensemble learning with structured and textual data

    Huang, R. et al. (incl. Wan, T. K.) · Computers in Biology and Medicine, 155, 106176 · preprocessing, NLP

  3. BMC Digital Health · 2023

    Machine learning-based prediction of COVID-19 mortality using immunological and metabolic biomarkers

    Tulu, T. W., Wan, T. K. et al. · BMC Digital Health, 1(1), 6

  4. Frontiers in Pharmacology · 2022

    Lung adenocarcinoma target gene prediction and drug repositioning

    Huang, R. et al. (incl. Wan, T. K.) · Frontiers in Pharmacology, 13

  5. Human Genetics and Genomics Advances · 2022

    Mendelian randomization-based exploration of RBC width and MCV with risk of hemorrhagic strokes

    Liu, J. et al. (incl. Wan, T. K.) · HGG Advances, 3(4), 100135

Experience

Career path

  1. Huawei

    Senior R&D Software Engineer — AI Platform & Framework

    • Core upstream contributor to vLLM-Omni (#26 in 2026): multimodal inference features, fixes, and tests
    • Module Owner — Omni-Coordinator: distributed multi-stage serving, registry, heartbeat, unhealthy-instance detection
    • Load-balancer integration (round-robin / least-queue / random) across GPU workers
    • Model Owner (Qwen-Image & related): deployment design, evaluation, release validation, production debugging
    • Mentored junior engineers and interns on distributed ML, benchmarking, and open-source workflows
  2. Hospital Authority

    Analyst Programmer / AI Research Officer — AI Systems

    • End-to-end healthcare AI: requirements → data → modelling → deployment
    • Parasite detection/classification (YOLOv5, EfficientDet, EfficientNet)
    • HCC prediction; human pose (MediaPipe, Detectron2)
    • Fine-tuned and deployed LLM chatbots; stakeholder liaison and reporting
  3. CityU · Biomedical Science

    Postdoctoral research

    • ML / statistical modelling on biomedical data, reproducible pipelines, interdisciplinary collaboration
  4. HKU SPACE

    AI & Robotics teaching (ApL)

    • Project-based teaching in ML, computer vision, and robotics; mentoring students building and debugging AI models and autonomous systems
  5. CityU · EE3211

    Teaching Assistant — Modelling Techniques

    • Supported teaching of statistical modelling and R (~50 students/class): hypothesis testing, regression/GLM, model comparison, and validation
  6. Education

    City University of Hong Kong

    • Ph.D. Electrical Engineering (2019–2023) — applied biomedical / health data science
    • BEng Computer Engineering (2015–2018) — First Class Honours · Dean’s List · CGPA 3.65

Open Source

vLLM / vLLM-Omni

Core contributor and Omni-Coordinator module owner — distributed coordination for multimodal, data-parallel GenAI serving.

#26 contributor on vLLM-Omni (2026)

Module Owner — Omni-Coordinator

Load balancer — routing policies across GPU workers

View GitHub →

Awards

Scholarships

Competitive academic and entrance awards during undergraduate and postgraduate study at City University of Hong Kong.

  • HKPFS Academic Excellence and QE Award

  • Chow Yei Ching School of Graduate Studies Entrance Scholarship

    Research Entrance Scholarship

  • Universal P.C.B. Equipment Co., Ltd. Scholarship

  • HKSAR Government Scholarship

  • Excellent Academic Performance Scholarship for CSE Students

  • The Hong Kong Real Property Federation Scholarship

  • HSBC Hong Kong Scholarship

  • Universal P.C.B. Equipment Co., Ltd. Scholarship

  • HKSAR Government Scholarship Fund — Reaching Out Award

    Non-academic Award

  • Jockey Club Harmony Hall Scholarship

    Hall Scholarship

Skills

Stack

GenAI / Serving

vLLM · vLLM-Omni · LLM fine-tuning · RAG / chatbot · multimodal serving

Systems

Distributed serving · load balancing · multi-GPU orchestration · benchmarking

ML / CV

PyTorch · TensorFlow · Hugging Face · YOLOv5 · EfficientDet · MediaPipe · Detectron2

Languages

Python · R · C++ · Java · SQL · Matlab

Contact

Let’s talk

Open to conversations on GenAI platforms, open-source collaboration, or research.

wantszkin2003@gmail.com +852 9601 2141 github.com/NumberWan LinkedIn

Hong Kong