Quantitative research

I am a quantitative researcher and systematic trader at Qrigin Capital, a startup proprietary trading firm. I work on forecasting, execution, market making and live risk for books traded on OKX, Binance and Coinbase.

Resume (PDF) Email

Experience

Quantitative Researcher and Systematic Trader

Aug 2025–present (full time to May 2026, part time since) Qrigin Capital · startup proprietary trading firm, $35M · Remote

Alpha discovery: built an LSTM forecasting engine on phase-space features with full MLOps in MLflow, and integrated more than 200 multi-frequency factors with decay analysis and cross-sectional ranking, lifting win rates by 3.1 percent.
Directional strategy: developed a high-frequency taker strategy on tick data that captures short reversions with a confidence signal, deployed live at a post-fee Sharpe ratio of 8.2 over three months on Coinbase International.
Optimal execution: architected TWAP, VWAP and implementation-shortfall algorithms for CTA books, with microstructure timing overlays on child-order aggressiveness, cutting slippage by about 2 bps against TWAP across major exchanges.
Market making: built a dual-engine tick backtest stack with vectorized L2 queue-position fills, and developed and validated a grid-based market-making strategy with an out-of-sample Sharpe ratio of 9 over three months.
Trading and risk: ran CTA books on OKX, Binance and Coinbase, owning P&L, TCA and fee analysis. Managed live risk on $35M with leverage, position and beta limits, and trimmed exposure ahead of the October 2025 crash.
Led a two-person research team, mentored junior researchers, and reviewed methodology and code before deployment.

Python (PyTorch, MLflow, Polars, pandas, NumPy)CEX APIs (REST and WebSocket)GrafanaLinuxGit

Quantitative Researcher

Feb 2025–Aug 2025 InsightCheck Investments · systematic equity fund with CTA desk, $300M AUM · Shenzhen, China

Statistical arbitrage: developed a market-neutral long-short equity strategy combining PCA statistical factors and machine-learning signals with a covariance-based risk model, backtested at a post-cost Sharpe ratio of 2.4.
Factor research: researched fundamental and microstructure alpha factors for Chinese A-shares, validating more than 62 intraday factors with daily IC above 0.01, and ran cross-sectional portfolio optimization in Polars.
Order-book engine and infrastructure: solely engineered an L3 market-by-order reconstruction engine, prototyped in Python and migrated to C++, that processes the full A-share universe in about one second per day. Built a Dask backtester that cut evaluation cycles tenfold.

Python (Polars, Dask)C++SQLLinuxGit

Projects

2026
IMC Prosperity 4 Trading Competition
Ranked in the top 3 percent of more than 6,000 teams using market making and cross-asset arbitrage driven by counterparty fingerprinting. Python (PyTorch) and Rust.
2025
Crypto market maker with backtest engine
Built and deployed a live ETH market-making framework on OKX, combining EWMA flow alphas with a CNN that captures order-book patterns through phase-space features. It averaged 0.5 to 1 percent daily return, and the project led to my recruitment at Qrigin.

Code

market-making-lab
Event-driven market-making research lab for testing order-book signals, latency, queue assumptions, inventory controls and execution costs with reproducible synthetic examples. github.com/denniswu28/market-making-lab

Case studies and writing

Trading Systems and Execution
Five inspectable views connect a published LOB architecture, observed market-making and execution artifacts, microstructure equations, and a public-data BTC risk reconstruction.
Reading Execution Evidence Without Inventing a Benchmark
A compact guide to separating observed execution measurements from benchmarks that the stored dataset cannot support.
Small-account equities record: August–November 2025
A privacy-reviewed audit of a historical account record, including the observed equity path, drawdown, evidence gaps, and limits on attribution.

Case studies use public, synthetic, delayed or aggregated data. Historical research is not investment advice.

Methods

Python (PyTorch, Polars, pandas, NumPy, Numba, Dask, MLflow), Rust, C and C++, SQL and Unix. Markets: digital assets, commodity futures, global equities, options (Akuna 101 and 201) and prediction markets.