AI Engineer based in Hong Kong with a background spanning econometrics, academic ML research, and production AI systems. I design and ship end-to-end solutions — from model development and REST API design to cloud deployments on GCP and Azure, platform integrations, and in-house SaaS development. Across insurance, luxury retail, and ad-tech, I've led data platform transformations, built multi-agent AI systems, and driven AI initiatives across APAC. I care about systems that are scalable, maintainable, and actually move the needle.
Training a reinforcement learning agent to play Blackjack optimally using policy evaluation, Monte Carlo methods, and value iteration — from coursework research at CityU.
Digital transformation & ML infrastructure for enterprise clients across insurance and media.
Led pipeline development and digital transformation for Prudential and BLP — building in-house ETL pipelines, migrating legacy SaaS systems to M Query and DEX, and shipping MLOps dashboards. Ran POCs on Azure, Databricks, and ML Studio to establish scalable ML infrastructure.
Led all AI initiatives across APAC retail — scalable data products, recommendation engines, and LLM agents.
Designed multiple scalable data products that cut overhead and drove €3M+ in incremental APAC revenue. Led the codebase transformation of the data platform and spearheaded nearly all AI initiatives across the APAC data team in Hong Kong — recommendation systems, forecasting models, federated learning, and LLM agents.
Building AI solutions for APAC insurers — reducing manual workloads and modernising the policy lifecycle.
Building AI solutions for APAC insurers — reducing manual workloads, improving straight-through processing (STP), and integrating AI across the policy lifecycle via agents, cloud platforms, and A2A orchestration.
Papers
Research I've contributed to.
Projects
Things I've built.
AI-powered native iOS app (Dart) helping kids learn through play. FastAPI backend on GCP with Firebase auth and custom AI agents for OCR content recognition. Secured HKD 100K seed funding from Cyberport CMF.
Original research and implementation of federated learning — distributed ML across decentralized devices without sharing raw data. Implements FedAvg and FedReconstruct with privacy analysis.
RL agent learning to play Blackjack optimally using policy evaluation, value iteration, and Monte Carlo methods. SDSC 6007 coursework.
Collaborative filtering and content-based recommendation engine from scratch. Matrix factorization and item-based similarity models.
Web application for flight planning — handles routing, scheduling, and data lookup with a clean Python API layer.
NLP classifier predicting product ratings from text reviews using TF-IDF, embeddings, and gradient boosting.
Interactive data visualization dashboard with R Markdown — analytical storytelling with real datasets and Shiny interactivity.
Blackjack RL Agent
Reinforcement learning agent learning to play Blackjack optimally.
Work History
Where I've applied and built.
AI Engineer
- Built end-to-end AI Underwriting Agent for Life Insurance on Azure with CI/CD and Docker
- Architected Agent-to-Agent (A2A) orchestration for tool centralization and AI Agent Builder platforms
- Built Siamese neural networks for forgery detection and wet signature verification
- Developed cross-product sales intelligence tool using multi-agent workflows (orchestrator, evaluator, aggregator, domain agents)
- Led ML/LLMOps workflows including ETL pipelines and monitoring on Databricks and Power BI
Co-owner & Lead Developer
- Secured HKD 100K seed funding from Cyberport Creative Micro Fund
- End-to-end development of native iOS app (Dart) and FastAPI backend deployed on GCP
- Integrated custom AI agents for OCR-based content recognition with Firebase and MySQL auth
Regional Data Scientist
- Delivered propensity-model pipelines (XGB/RF/LGBM) driving €3M+ incremental product sales (FY25–26)
- Deployed Deep Matrix Factorization and GRU models with 68–82% Top-K accuracy for product recommendations
- Productionized 14 boutique performance forecasting models at 60–85% monthly accuracy
- Deployed product-recommendation AI agent powered by Chainlit, LangGraph, ChromaDB, MCPs, and Redis
- Implemented Variational Autoencoders and Federated Learning to mask PIIs for model training
Data Scientist
- Fine-tuned and deployed a Content Classification Model (BERT) at 66% accuracy using FastAPI and Docker
- Built fraud detection models for insurance claims using Keras and Azure ML Studio (70% Recall)
- Developed in-house RAG chatbots with graph generation using FAISS, PaLM, LangChain, and Looker SDK
- Python API developer for in-house AI marketing content generator (Django, Gemini, RAG)
Research Analyst
- Migrated legacy SQL scripts to R, optimizing data workflows with custom geocoding functions
- Location analysis with spatial data, geocoding, and construction funding estimates using R and SQL
Project Assistant / Learning Developer
- Designed online courses and vector arts using Articulate 360, WordPress, and Adobe Creative Cloud
Event Management Trainee
- Monitored membership database and produced P&L reports using Excel and NeonCRM
- Designed infographics, booklets, and marketing campaigns using Adobe CC and Mailchimp
Academic Background
Degrees and certifications.
Master of Data Science (MSDS)
Bachelor of Arts — Econometrics & Business (Co-op)
Professional Machine Learning Engineer (PMLE)
Get In Touch
Open to AI/ML projects, research collaborations, and opportunities.