Spencer's Projects

Optimal High-Frequency Market Making | Python · WRDS TAQ · Market Microstructure · Stochastic Control

Implemented and extended the Avellaneda-Stoikov (2008) market making model with a proprietary event-driven backtester calibrated on real WRDS TAQ data across five equities. Contributions include independent κ calibration via exponential regression, empirical volatility estimation, and dual queue-priority modeling. Optimal strategy improved P&L over the NBBO baseline across all five tickers while significantly reducing end-of-day inventory variance. Columbia University · IEOR 4733 · with G. Naples, D. Niedfeldt, B. Zenker

Algorithmic Trading via Mean Field Games | Python · SDEs · Price Impact · Market Microstructure

Built a mean-field game simulation framework for optimal execution with heterogeneous agents, latent market regimes, and price impact dynamics. Models agent interactions via coupled Hamilton-Jacobi-Bellman equations and Fokker-Planck distributions, extending the Casgrain-Jaimungal framework to multi-population settings with posterior belief filtering.

Output Gap Estimation via Particle Filter | Python · NumPy · SciPy · Monte Carlo · State-Space Modeling

Simulated output gap dynamics via an AR(1) linear Gaussian state-space model, comparing maximum likelihood estimation, Kalman filtering, and sequential Monte Carlo particle filtering across 1,000 reproducible time steps. Evaluated posterior accuracy and computational tradeoffs across all three methods. Columbia University · with Aidan Lowe

Bloomberg Trading Competition | Bloomberg Terminal · Risk Management · Equities

Ranked 1st in P&L among all Columbia University teams in the Bloomberg Trading Competition, managing a live simulated portfolio with real-time execution timing, liquidity considerations, and intraday risk exposure across equity markets.