I study where and when information enters prices, and what it costs the market maker who is quoting.

Optiver × UChicago Explaining Markets

Ranked first globally; final results pending. A live earnings-reaction model explained an extra 0.150 of R squared beyond the earnings surprise on 517 events (95% CI 0.10-0.20), 27% above the strongest LLM baseline. Every model change had to pass out-of-sample replication, which cut one fix's estimated gain ~60% and rejected two others; the live LLM pipeline processed 584 of 584 events with zero errors (~1 s median latency per event).

IMC Prosperity 4

256th of 18,803 worldwide (top 1.4%). Market making around Ornstein–Uhlenbeck fair values, with quotes skewed on book imbalance to limit adverse selection; on a 10-strike call chain the edge was only near the money.

Order-level liquidity around the EIA releases

A pre-registered study of 60 EIA inventory releases (March to October 2026, 440 million CME order-book records), each against four matched control days, with the analysis frozen before data access. After the release, depth within 5 ticks of the mid stays depleted for the equivalent of an extra 34 seconds in crude and 48 seconds in natural gas of fully missing depth over the first five minutes; about a third of the crude effect also appears at 10:30 on gas-report days. Before the release, the book thins at the best price because fewer new orders arrive in crude (about 60% of that also before the gas report) and more orders execute in gas, not because traders cancel more: contrary to my initial hypothesis. Failed placebos are reported alongside.

Status. Confirmatory analysis done and its primary tests independently replicated. Out-of-sample replication on releases from mid-October 2026 in progress.

How wrong are backtest queue models?

Exact FIFO queues rebuilt from CME order IDs, used to score the queue-position estimators that market-making backtests rely on when they only see aggregate depth: queue error, fill-rate bias, and how much of the P&L error comes from adverse selection. C++20 core with Python bindings.

Status. Library built and tested on synthetic data with known answers, including bitwise agreement between two independent implementations. Real-data runs next.

C++ limit order book and market-making simulator

A deterministic single-threaded matching engine (price-time priority, partial fills, cancel/replace, self-trade prevention) at about 3.7M events per second and 125 ns median latency on an Apple M3, with 69 tests and CI. In simulation under order flow calibrated to Nasdaq ITCH data for QQQ, quote fill rates fell from about 60–74% to under 2%; the diagnostics trace this to sparse executions rather than queue position.

Cross-venue price discovery in BTC-USD

Coinbase leads Kraken at short horizons: Granger-significant after Benjamini–Hochberg correction in 6 of 9 usable half-hour sessions, against 1 of 9 in reverse, on quotes aligned to a 100 ms grid. Long-run leadership varied by session.

Honors thesis: Bayesian sequential decision-making under regime switches

Studied Bayesian sequential decision-making under regime switches: an HMM regime detector with a CUSUM change-point trigger feeding Kelly position sizing, evaluated by Monte Carlo on detection delay, ruin and median wealth.

WH Trading, University of Chicago Project Lab

Forecasting realized volatility across horizons in futures markets against implied volatility, classifying momentum and mean-reversion regimes, and testing whether regime-aware delta hedging beats fixed-threshold rules on P&L, risk and turnover.

Quantitative Modeling Intern, Endless Moments

ARIMA and LSTM time-series forecasts, validated walk-forward.

Quantitative Research Intern, Morrison School of Agribusiness (USDA-funded)

A Box-Cox test that reproduces OLS exactly at λ = 1, with Jacobian-corrected likelihood profiling in Python and Stata/Mata.

Payments Automation Intern, American Airlines

SQL and Python settlement-validation checks for payment reconciliation.