Siddhant Shah

Studying the market microstructure of DeFi perpetual futures — instruments that processed $41B+ in daily notional volume by early 2026 — across crypto, tokenized equities, and tokenized commodities. - Constructed a novel multi-asset, multi-exchange dataset covering 17 assets (5 crypto, 8 tokenized U.S. equities, 4 tokenized commodities) across 3 DeFi platforms over a 7.5-month window, ingested via public REST APIs and Yahoo Finance. - Developed an endogenous event-detection methodology using rolling 3-day t-tests to identify abnormal trading volume without a predetermined event calendar — adapting classical event study methods to 24/7 DeFi markets with no opening bell. - Identified multiple statistically significant volume anomalies across 17 assets and documented their macroeconomic catalysts: the U.S. Crypto Strategic Reserve announcement, a Fed independence crisis driving, and a 15-of-17 asset co-exceedance on FOMC + mega-cap earnings day. - Quantified a novel "24/7 DeFi premium" market-structure effect: systematic Saturday volume crashes (thin liquidity, no TradFi price anchor) and holiday volume spikes — a dynamic with no direct CeFi analogue. - Measured price tracking accuracy across asset classes using Pearson correlation and tracking error, documenting a sharp maturity gradient.

Quantitative finance research spanning portfolio optimization, systematic trading strategy development, cross-asset signal research, and ensemble ML methods — resulting in 3 published papers. - Designed and implemented a comparative study of mean-variance vs. robust dispersion-constrained portfolio optimization (MAD, IQR constraints) across 3 risk budgets and 4 rebalancing frequencies using 25+ years of daily data on 9 U.S. sector ETFs. - Backtested momentum-based crude oil strategies across ETFs and futures over 18 years, generating 10.6% ETF / 19.9% futures annualized returns vs. −8.6% buy-and-hold; published in Stocks & Commodities, Sept 2025. - Engineered a cross-asset Bitcoin/silver model using overnight silver returns as a directional signal for Bitcoin 24-hour moves — delivering 68.65% average annual returns with ~34% lower drawdowns than buy-and-hold over 10 years; published in Stocks & Commodities, June 2025. - Developed a probabilistic ensemble accuracy framework using Poisson, Normal, Binomial, and Poisson-Binomial approximations, reducing model optimization time ~80% and achieving <3% mean relative error; published in MLAIJ, May 2025. - Built modular Python backtesting infrastructure with strict no-look-ahead enforcement, rolling-window estimation, and proportional transaction cost modeling to ensure realistic strategy evaluation. - Produced standardized performance and risk analytics (Sharpe, max drawdown, turnover, tracking error) across strategies and rebalancing schemes, packaged into reproducible notebooks for research and internal reporting.

Supported instruction for a graduate-level data analytics course, leading office hours and reviewing student work on machine learning and data pipeline projects.

- Led data-driven expansion analysis for the Maths Circle Initiative, using Python and Excel to consolidate applicant data and build dashboards on geography, conversion rates, and program capacity — informing regional deployment decisions. - Implemented project management infrastructure (Jira, Confluence, Slack) and improved cross-functional coordination across the foundation's program teams.

Prepared high-achieving students for the International Junior Science Olympiad, with emphasis on quantitative reasoning, probability, and data interpretation.

Built the bank's big data analytics infrastructure for government-sponsored financial scheme analysis and customer behavior intelligence. - Engineered an end-to-end PySpark and SQL analytics pipeline processing 6,000+ government-sponsored financial schemes and customer transaction datasets at scale — the bank's first structured pipeline for this data. - Designed pattern recognition algorithms on customer-level transaction data to extract behavioral segments and cross-sell signals, producing clean analytical tables and segment-level reports for senior leadership. - Developed a data-driven revenue growth strategy from pipeline outputs — projected to double revenue — by identifying underserved customer segments and high-yield cross-sell opportunities across scheme categories.

Acted as a Teaching Assistant for the Design and Analysis of Algorithms (ALGO) course for M. Sc. Data Science students.

Acted as a Teaching Assistant for the Classical Mechanics 1 (CM1) course for B. Sc. Undergraduate students