I am a fifth-year Ph.D. candidate in Operations Research at Columbia University, advised by Professor Agostino Capponi.
Broadly, I study the design and operations of digital platforms and financial technologies. Within this agenda, my current research interests focus on prediction markets and information aggregation, tokenization and its economic consequences, and AI-enabled financial decision-making.
My research has been presented and published at top conferences in economics, finance, operations research, and computer science. I've also collaborated with leading industry research teams on stablecoin market analysis and decentralized exchange design.
Latency competition in blockchains results in transaction spamming. Auctions for artificial time priority reduces this spam and raise platform revenue. We verify these results using data from Arbitrum’s Timeboost launch as a natural experiment.
Prices in prediction markets converge to 0 or 1 by a fixed time, resulting in large arbitrage losses to liquidity providers under typical AMM settings. Proper design of AMM invariants and dynamic liquidity schedules can smooth this loss across both price and time.
Revert protection waives gas fees for failed transactions. On blockchains using priority gas auctions, this improves auction revenue, price discovery, and blockspace efficiency.
Market depth on DEXs can be decomposed into capital deployment and capital efficiency channels. We measure how volatility, fees, adverse selection, gas costs, and external liquidity affect each channel.
We build an agentic AI pipeline that clusters prediction markets and identifies correlated or contradictory markets based on the their textual characteristics, then construct semantic-based trading strategies from AI-identified relations.
We develop a framework for defining and evaluating privacy enhancements in payment systems, then use it to design systems balancing legitimate privacy benefits against illicit activity, and to compare existing systems.
Passive market makers benefit from active market makers in DEXs due to the pro-rata sharing of trading fees, but can get crowded out when trading volume is insufficiently responsive to liquidity, worsening market quality.
Liquidity pool states in AMMs can be represented as order books, allowing for optimal execution and liquidity allocation in AMMs to be analyzed via order book perspectives.
We examine the technological, legal, and regulatory implications of tokenizing real-world assets.
Water trapped inside sinking tectonic plates can trigger earthquakes deep below Earth’s surface. Using unsupervised learning, we isolate a thin zone where pressurized water transforms and fractures minerals, offering clues to the mystery of how earthquakes can still occur under intense pressures where ordinary rock should not break.