Yuheng (Paul) Yan

• Core Courses: Stochastic Process; Machine Learning; Investment Science; Advanced Statistical Theory; Advanced Equity Derivatives; Time Series Analysis; Interest Rate and Credit Derivatives; Commodity Markets; Mathematical Game Theory; Introduction to Data Science; Machine Learning in Finance

• Managed a 5-person quant team; owning an end-to-end trading system and reporting directly to CEO.
• Built the research→execution pipeline (data → signal → risk → backtest → paper/live) with robust monitoring under real-world constraints.
• Developed RL-based crypto strategies achieving Sharpe 1.95 and +85.3% out-of-sample return; improved generalization via reward/risk shaping and training stability.
• Directed NLP-driven sentiment analysis on crypto news and social media from different data sources, integrating signals to strengthen alpha generation.
• Designed investor-facing performance dashboard (PnL, exposure, drawdown, turnover) and collaborated with marketing to refine website messaging and product Q&A.
• Supported CEO in fundraising workflows by translating strategy/system capabilities into clear narratives, metrics, and diligence-ready materials.
• Validated internally developed strategies with personal capital, achieving a 100% return (2k → 4k) over two months.

• Engineered high-frequency features from Level II order book data—including order flow imbalance and volume clustering—to uncover microstructure-driven alpha signals.
• Applied LSTM model to high-frequency order book data, enhancing signal extraction and predictability precision.

Identifying and Quantifying Financial Bubbles with the Hyped Log-Periodic Power Law Model

Co-First Author | Supervisor: Prof. Helyette Geman | Preprint: arxiv.org/abs/2510.10878 | Presented at The 21st Quantitative Finance Conference 2025, Italy; Upcoming Presentation at QuantMinds International 2025, London.

• Co-developed a unified bubble detection framework integrating LPPL residual dynamics with NLP-based Hype Index and sentiment scores, leveraging behavioral finance to quantify overbought and oversold phenomena.
• Designed a dual-stream Transformer to jointly process stock-level and market-level signals, achieving MSE = 0.087 and Pearson correlation 0.625 in Bubble Index prediction.
• Implemented systematic trading strategies; delivered 34.1% average annualized return and Sharpe 1.13, with best cases exceeding 100% annualized return (Sharpe >3) in backtests for US market.

Systematic Arbitrage & Factor Structure Analysis

Research Assistant, Carey Business School, Johns Hopkins University, Jan 2025 – July 2025

• Replicated and extended a PPCA-based equity arbitrage framework from academic literature, constructing long–short portfolios; improved benchmark results by raising Sharpe from 0.7 to 1.6, annualized return from 18% to 23% and reducing max drawdown from 23% to 14%.
• Applied LASSO regression to high-frequency market microstructure features, isolating predictive signals while reducing dimensionality under real-time constraints.
• Benchmarked machine learning forecasts of quarterly P/E ratios; achieved MSE = 0.084.

• Guided students in building machine-learning pipelines for high-frequency trading using live tick data.
• Introduced NLP-based sentiment models (FinBERT/Vader) for portfolio optimization tasks.

• Supported crypto trading labs—on-chain flow auditing, stat-arb, RL trading, and AMM microstructure/slippage—helping students translate market microstructure + blockchain data into executable strategies.