Resume
Education
University of Melbourne
Ph.D. in Electrical and Electronics Engineering
2019.03 - 2023.08- Supervisors: Jingge Zhu, Jonathan H. Manton, Uwe Aickelin
- Research Topic: Information-Theoretic Analysis of Transfer Learning and Machine Learning Theory
- Published 10+ papers in top-tier venues including IEEE Transactions on Information Theory, JMLR and Artificial Intelligence (AIJ)
- Awarded Australian Government Defense Project for outstanding research performance
Master of Electrical and Electronics Engineering
2016.09 - 2018.12- Dean's Honours List 2019 (Top 3% among 500+ graduate students)
- Dean's Honours List 2018 (Top 3% among 500+ graduate students)
- GPA: 87/100 (Highest Distinction Average)
- Thesis: 3D Scene Reconstruction with Low-cost LiDAR - from Hardware to Software
Fudan University
Bachelor of Electronics Engineering
2012.09 - 2016.06- GPA: 3.22/4 (Top 30% in a highly competitive program)
- Thesis: Image Segmentation and Fusion Algorithm for Pseudo-HDR
- Received National Scholarship (Third Class), 2015-2016
Work & Research Experience
Heidi Health
Staff AI Engineer
2026.08 - Present- Tech lead for an air-gapped, on-premises deployment of Heidi's clinical AI stack (transcription, note generation, dictation, translation and evidence retrieval) for hospitals with data-sovereignty requirements, served entirely on local NVIDIA GPUs across 8 languages
- Distil multilingual clinical ASR from commercial teachers into self-hosted Whisper and Qwen students: the Korean model reached parity with the commercial cloud engine in blind clinician review, and a single Arabic/Urdu/English model outperformed its teacher on clinical Arabic
Senior AI Engineer
2025.10 - 2026.08- Delivered fine-tuned LLMs for clinical note generation (SFT, DPO, distillation) that replaced expensive closed models in production, saving Heidi millions of dollars per year
- Led the transcription workstream: fine-tuned an in-house clinical ASR model (Parakeet) that took over most English production traffic from vendor APIs at lower latency and cost
AI Engineer
2025.06 - 2025.10- Built the post-training data pipeline for clinical note generation: sampled production sessions across five regions, regenerated target notes and filtered them with LLM judges before SFT and DPO
- Ran side-by-side experiments between fine-tuned variants and frontier models on live traffic, tracking clinician satisfaction and edit rates
- Shipped an improved clinical code generation model, and contributed to a collaboration with NVIDIA on fine-tuning speech recognition for clinical use
StrongRoom AI
Machine Learning Engineer
2023.06 - 2025.05- Spearheaded the architecture and implementation of StrongPro Platform, a healthcare analytics solution integrating ML with EHR systems
- Enhanced pharmacy-patient interactions and adherence tracking across 400+ pharmacies, generating AUD 65M+ revenue with AUD 1.5M attributed to improved algorithms
Data Analyst
2022.12 - 2023.06- Developed the StrongPro Platform for pharmacy owners and medical institutions, targeting patient adherence and loyalty prediction
- Analysed and visualised medical data to surface insights on patient behaviour, pharmacy revenue, and recommended interventions
- Built the initial NLP and ML pipelines for medical data extraction and adherence computation
University of Melbourne
Research Fellow, Applied & Multi-modal AI
2024.01 - 2025.02- Built an interactive framework for evaluating L2 English learners with LLMs; accepted at NAACL HLT 2025 (oral, top 15%)
- Analysed how faithfully LLMs simulate L2 learners across different L1 backgrounds; accepted to ACL 2025 Main Conference
- Developed real-time multi-modal out-of-distribution detection for long dialogues, combining text and visual signals; accepted at ACL Findings 2025
Teaching Fellow
2019.04 - 2024.12- Head Tutor for ELEN90052 Advanced Signal Processing; ran labs and tutorials for 50+ graduate students per semester
- Taught ELEN90058 Signal Processing (2021-2022) and wrote the tutorial material for both courses
- Ranked first in the department for teaching satisfaction (2019); supervised final-year projects, two students went on to Ph.D. study
Data Analyst Researcher
2019.03 - 2019.12- Developed the machine learning algorithms behind the Fertility After Cancer Predictor (FoRECAsT), reaching state-of-the-art prediction performance
- Deployed across three major hospitals; results published in healthcare journals and book chapters
Technical Skills
- AI / ML
- LLM post-training (SFT, DPO, LoRA, distillation)ASR (CTC, TDT, Whisper, Qwen-ASR)Multi-modal modelsEvaluation & benchmarking
- Programming
- PythonPyTorchHugging FaceRustMATLABJavaScript / VueLaTeX
- Infrastructure
- DockerKubernetesAWSAirflowSparkFastAPI
- Model serving
- vLLMSGLangNVIDIA DynamoTensorRTTriton Inference ServerAir-gapped deployment
- Databases
- MongoDBPostgreSQLMySQLDelta Lake
- Languages
- Chinese (native)English (proficient)
Key Achievements
- 10+Papers in top venues (IEEE TIT, JMLR, ACL, NAACL), h-index 7
- 2023Best EEE Ph.D. Thesis Prize, University of Melbourne
- AUD 30k+Research funding and project grants secured
- IEEE TITOutstanding Reviewer; also reviews for ISIT, ICASSP, TMLR, ICML
- AUD 1.5M+Annual revenue improvement from ML systems I built
- MillionsUSD saved per year by replacing closed models in production with fine-tuned ones