About Me

I am a Computer Science undergraduate at The University of Hong Kong, pursuing a second major in Finance. I maintain a CGPA of 3.73/4.30 and have received both the Dean's Honour List and the C. V. Starr Scholarship.

My interests are in machine learning, data science, and optimization. I have hands-on experience across data governance, feature engineering, time-series backtesting, model development, and AWS deployment, alongside finance and Web3 projects.

I am particularly interested in ML engineering, optimization, and FinTech applications. If you would like to connect or discuss a project, please feel free to reach out.

Skills
Programming
Python C / C++ Java HTML/CSS/JS PHP SQL Solidity Bash
Data & Machine Learning
PyTorch scikit-learn Pandas / NumPy Matplotlib LightGBM / XGBoost SHAP / GeoPandas
Tools & Environments
Git Docker WSL 2 Poetry / Pytest AWS S3 / Redshift AI Agents / RAG
Languages
Mandarin (Native) English (IELTS 7.0) Cantonese (Basic)
Education
Sep 2024 – Jun 2028
The University of Hong Kong
BEng in Computer Science · Second Major in Finance
CGPA: 3.73/4.30 · Dean's Honour List (2024-2025) · C. V. Starr Scholarship (2025-2026)
Jan 2027 — Jun 2027
National University of Singapore
HKUWW Exchange Program
Experience
Jun 2026 – Sep 2026
Data Science Intern · Oxbridge Economics (Hong Kong)
Developed an end-to-end ML forecasting solution for Annual Bluegrass Weevil life cycles: governed observation and weather data, refactored legacy experiments into a reproducible configuration-driven pipeline, and advanced the production model to LightGBM with Ridge termination calibration. Built time-series backtests, diagnostics, technical documentation, and an AWS deployment workflow.
Jul 2026 – Sep 2026
AI Model Engineer Intern · Huawei Technologies, Beijing Research Institute
Developed, tested, tuned, and deployed optimization algorithms for customer objectives. Applied feature engineering and clustering for dimensionality reduction; implemented Greedy, Local Search, and ALNS heuristics; and contributed example code and market research for an AI solver coding agent and CP/SMT solver ecosystem.
Jun 2025 — Jul 2025
Financial Intern - CITIC
Systematically studied fundamental financial products and their operating fundamentals.Undertook hands-on practice in basic tasks such as stock management, asset allocation, and wealth management. Paid a visit to the company’s headquarters in Beijing to gain an understanding of the organizational structure and primary functions of departments within the securities industry. Visited a range of public and private fund companies, engaging in exchanges and learning activities with fund managers.
Oct 2024 — Oct 2025
Internal Secretary
Chinese Scholar and Student Association (CSSAUD)
Projects
WeevilTrack Turf Pest Phenology Forecasting System
Machine Learning & Data Science · Oxbridge Economics
Refactored and productionized an end-to-end ML system for forecasting Annual Bluegrass Weevil life cycles, covering data processing, feature engineering, training, prediction, backtesting, and evaluation.
  • Upgraded the core model from Random Forest to LightGBM, adding signed-time targets, GDD/CDD, rolling weather features, monotonic constraints, and post-event samples.
  • Designed a LightGBM base-forecast plus Ridge termination-calibration architecture. In controlled backtests over 228 historical trajectories and 10,999 forecasts, reduced MAE from 8.35 to 4.73 days and RMSE from 12.00 to 6.67 days; improved the within-five-day hit rate from 48.6% to 69.2%.
  • Built rolling time-series validation and diagnostics for 2022–2025 using feature ablation, SHAP, partial-dependence plots, and cross-region holdouts; prepared versioned AWS S3 releases for daily incremental forecasts and safe rollback.
PythonLightGBMscikit-learnSHAPGeoPandasAWS S3Redshift
View GitHub repository ↗
Web3.0 Quantitative Trading Bot (HackerU Trading)
Lead Strategy Developer & Data Feature Engineering
Built an automated API-based data ingestion pipeline and an order-flow-driven feature engineering module for high- and medium-frequency quantitative strategies.
  • Engineered OHLCV and aggressive buy/sell volume features, including active buy-sell ratio, order-flow imbalance, rolling net aggressive buying, volume intensity, VWAP deviation, order-flow acceleration, return Z-score, and wick extreme ratio.
  • Implemented rolling-window alignment and future-data isolation to reduce look-ahead bias across training and backtesting.
  • Built a LightGBM + XGBoost ensemble with a multi-layer trading funnel: meme/L1/AI sector constraints, BTC trend switch, RSI overheating filter, ATR volatility threshold, Edge ranking, losing-streak cooldown, and gap handling.
  • Designed dynamic position management with ATR-adaptive take-profit/stop-loss rules (2.0-2.8x ATR) and a 12-bar maximum holding period. Historical data was pulled concurrently, cached, and used for daily rolling retraining; 10-day backtests reached up to 1% ROI.
PythonNumPyPandasscikit-learnLightGBMXGBoost
View GitHub repository ↗
HKU AGENTS · Agentic Knowledge and Campus Assistant
Agent Architecture & Full-Stack Development
Built an HKU-focused assistant that combines official-source retrieval with campus tools for Portal, SIS, Moodle, and Library workflows.
  • Combined parent–child chunking, Qdrant dense retrieval, and BM25 to support bilingual clarification and evidence-backed answers across multiple LLM providers.
  • Built a local-first platform with a TypeScript DeepSeek Harness plugin, versioned FastAPI tools, and a restricted Chrome Bridge that reuses authenticated sessions while protecting credentials and private page data.
  • Designed a controlled Library booking flow with exact-slot previews, explicit confirmation, policy checks, and post-submit verification. Read-only workflows have completed live acceptance; supervised booking awaits full live acceptance.
PythonLangGraphQdrantBM25FastAPITypeScript
View GitHub repository ↗
Web3.0 Stock Hedge Insurance Protocol
Smart Contract Development & Risk Pricing System Design
Designed and implemented a Solidity-based stock price insurance protocol where users purchase upside or downside protection for US/HK stocks with MockUSDC, while LPs provide stablecoin underwriting liquidity.
  • Developed core contracts including PolicyFactory, InsuranceVault, PricingOracle, ChainlinkOracleAdapter, and an automated settlement module.
  • Implemented policy purchase, cancellation, expiry settlement, LP share accounting, premium revenue, payout statistics, multi-asset whitelisting, market-hour restrictions, and Hong Kong market lunch-break handling.
  • Built an on-chain lightweight pricing model using volatility, direction, duration, trigger distance, pool utilization, and risk score.
  • Added multi-layer risk controls such as per-symbol underwriting caps, directional exposure limits, tenor limits, minimum liquidity buffers, oracle fallback checks, and pause controls. Used OpenZeppelin AccessControl, Pausable, ReentrancyGuard, and SafeERC20; completed unit, integration, fuzz, and stress tests with all Hardhat tests passing.
SolidityHardhatOpenZeppelinChainlinkDeFi
View GitHub repository ↗
C++ Game Development
Lead Developer (Course Project)
Designed and implemented the core combat module of a survivor game, including player controls, dynamic zombie generation, collision detection, and console-based UI rendering.
C++Game LogicConsole UI
View GitHub repository ↗
Quick Info
Year
Junior (Class of 2028)
Location
Hong Kong, HK
Research Focus
ML · FinTech · Quant Trading