Research
While much of my work remains proprietary, I publish select insights into my methods.
Market Microstructure
Liquidity dynamics, order flow analysis, and cross-venue arbitrage in fragmented markets.
ML & RL
Adaptive execution under constraints, regime detection, and policy optimization for trading.
MEV & DeFi
Quantifying extraction costs, sandwich attack patterns, and MEV-aware portfolio management.
Prediction Markets
Probability coherence, resolution mechanics, and structural inefficiencies in prediction venues.
Learning Under Adversarial Dynamics: MEV‑Aware Market Simulation and Multi‑Agent Reinforcement Learning
Current MARL research often relies on simplified games. I propose using MEV‑dominated blockchain markets as a high‑fidelity testbed for adversarial learning, policy convergence, and systemic fragility, bridging probabilistic regime modeling with strategic agent interaction. I plan to pursue a PhD centered on this agenda.
High‑Fidelity Adversarial Simulation
Build a mempool‑aware simulator that captures gas auctions, latency variability, and atomic sandwich/backrun mechanics to study equilibrium formation.
MARL Under Structural Constraints
Train heterogeneous agents (LPs, searchers, arbitrageurs) in non‑stationary, partially observable, delayed‑reward environments to analyze exploitation equilibria.
Regime‑Conditioned Policy Optimization
Build latent‑state observers from microstructure and on‑chain signals, conditioning policies on execution regimes (gas spikes, liquidity dry‑ups) to shift between aggressive and risk‑off behavior.
Market Mechanism Design
Use simulation data to identify tipping points where MEV incentives create systemic risk, then propose learning‑compatible rules to mitigate adversarial lock‑in.
Equilibrium Execution: Game-Theoretic Optimal Trading Under Endogenous MEV on AMMs
Modeling DEX execution as a Stackelberg game between traders and adaptive searchers, with artifact-backed evaluation across complexity regimes, calibration transfer, algorithm sensitivity, and out-of-distribution stress boundaries.
Hybrid Regime Detection and Risk Management in Semiconductor Equities: A Bayesian HMM-LSTM Framework
This research proposes a Bayesian HMM-LSTM framework to address high volatility and frequent regime shifts in semiconductor markets. By integrating probabilistic regime detection with long short-term memory networks, the model prevents inefficient capital allocation and provides robust risk management against geopolitical and macroeconomic shocks.
Interested in collaborating on research?
Get in Touch