CV

Education

• B.S. in Mathematics and Applied Mathematics, Fudan University, 2023.09 – 2027.06 (expected)
  • Selected coursework: Probability Theory (A), Mathematical Statistics, Deep Learning (A), Data Structures, Optimization Methods, Stochastic Analysis in Finance
• Exchange Student, Department of Mathematics, The Hong Kong University of Science and Technology (HKUST), Fall 2025

Experience

• Quantitative Research Intern, Quant Team, Soochow Securities, 2026.05 – Present
  • Converted existing cross-sectional stock-selection factors into time-series (timing) signals, improving factor adaptability across different market regimes.
  • Refined existing factors and evaluated their performance, and mined new Alpha factors from data characteristics.

Selected Projects

• factor_lab — A Reproducibility Framework for Published Equity Factors — An independent A-share factor research stack (point-in-time financials, back-adjusted prices, Barra CNE5 neutralisation, a 53-operator expression DSL, Newey–West IC inference, layered backtesting with costs) built to replicate factors published in sell-side research and test them under proper multiple-testing correction.
  • Pilot batch of 24 factors: 22/24 significant unadjusted and 22/24 surviving Benjamini–Hochberg FDR, but 0/24 surviving a Deflated Sharpe Ratio threshold of 0.95 — against an expected maximum Sharpe of ≈1.6 under the null across 24 trials.
  • Replicated performance fell short of published figures 93% of the time, median shortfall −5.77% annualised.
• Time-Series Momentum & Carry Factor Library — Built a two-factor-family research pipeline on ~15 years of daily data for 30+ Chinese commodity futures continuous contracts: time-series momentum (multi-window, volatility-normalised) and carry (term-structure slope), both standardised and sector-neutralised, with a full IC / ICIR / monotonicity / half-life evaluation suite.
  • Momentum portfolio: annualised Sharpe 1.15. Carry long–short: Sharpe 0.9 (IC 0.06), strictly monotonic across quintiles.
  • IC-IR weighted composite: annualised Sharpe 1.6, max drawdown held within 9%.
• Machine-Learning Factor Mining (GBDT) — Constructed a 60,000-row × 40+-dimension feature matrix from TA-Lib technical indicators, labelled with the triple-barrier method, and trained an end-to-end GBDT model under time-series cross-validation to eliminate label leakage.
  • Out-of-sample IC 0.06, directional accuracy 55%, long–short portfolio annualised Sharpe 0.95, annualised return 14%.
• Deep-Learning Time-Series Forecasting (LSTM / GRU) — Reused the same feature system to implement LSTM and GRU sequence models in PyTorch, with sliding-window sequence sampling and multi-model ensembling to better capture non-linear interactions.
  • Out-of-sample IC 0.08, directional accuracy 57%, long–short portfolio annualised Sharpe 1.45 (≈53% above GBDT), max drawdown narrowed to 9%.

See the portfolio page for longer write-ups.

Skills

• Programming: Python
• Libraries: PyTorch, NumPy, Pandas, Matplotlib, TA-Lib
• Tools: LaTeX, Markdown, Git
• AI agents: heavy Claude Code user
• Languages: Chinese (native), English (CET-4, CET-6)

Honors and Awards

• Fudan University Freshman Scholarship
• Basic Science Scholarship, Fudan University — 2023, 2024, 2025
• Chinese Physics Olympiad (CPhO), Third Prize — 2023
• Gaokao (National College Entrance Examination): 694, top 100 in province

Service and leadership

• Mid Lane Shotcaller in League of Legends (LoL) :)