I’m an Assistant Professor of Finance at the Kate Tiedemann School of Business and Finance, University of South Florida. I earned my Ph.D. in Finance at the University of Florida in 2023, and before that I studied economics and finance at the University of Alberta and at Korea University in Seoul.
My research brings machine learning to finance, and I’m especially drawn to measurement and benchmarking problems. A lot of what we treat as data is really an estimate: the fair value of an option, the price of a bond on a quiet day, the research a firm does but never reports. I believe better estimates can change what we think we know, and that idea ties together everything I work on.
Most of my work is in stocks, options, and bonds, within asset pricing and market microstructure. I’m curious about how prices get discovered, and what machine-speed trading costs the people on the other side of it. The same approach carries over to corporate finance and governance too, such as measuring innovation.
I like teaching programming, big data, and machine learning for finance. I also build open-source tools for quantitative research, mainly quantpolars for fast option pricing and fintechtools for containerized computing on HPC systems.
I’m grateful for a happy Christian family with three sons, and most of my time away from the desk goes to them.
Curriculum vitae (PDF)
GitHub — code and repositories
Research
Machine learning. I think machine learning helps finance most when it measures things, not only when it forecasts. Much of what we'd like to know isn't directly observed: what an option is really worth, or what a firm invents but never discloses. ML can be extremely powerful for estimating those quantities, and then we can ask which conclusions change.
Asset pricing — options, bonds, microstructure. Options and bonds are where this pays off first, because high-dimensional surfaces and thin trading are exactly where these methods shine. From there, I study trading at machine speed: who trades against option market makers, how much of it is a race on information that's already public, and who ends up paying for it through wider spreads.
Corporate innovation. Firms do a lot of R&D they never report, and ML can recover it from what they do report. Once innovation becomes measurable, our picture of corporate investment shifts with it. Governance and executive pay look like promising places for the same kind of measurement.
Working papers updated September 2026
High-Frequency Traders in the Options Market: Cream Skimming and Toxic Order Flow
Mahendrarajah Nimalendran and Matthew G. Son · Working paper
We develop a theory-based model separating toxic from non-toxic arbitrage in options markets and test it on high-frequency data. Fast traders systematically select against option market makers, and the resulting adverse selection shows up in spreads and quoted depth rather than in prices alone.
SSRN
What If Option Closing Prices Were Trustworthy? A Machine Learning Approach
Alejandro Lopez-Lira, Mahendrarajah Nimalendran, and Matthew G. Son · Working paper ASX Best Derivatives Paper Award
Options have no closing auction, so the market relies on last-trade prices that are stale and on closing quotes that are easy to distort. We estimate a counterfactual closing auction price with machine learning and find that 4 PM mid-quotes and last-trade prices deviate from it by 33 and 43 percent, which points to real efficiency gains from running a closing auction in options.
SSRN
Bond Price Discovery with Machine Learning
Mahendrarajah Nimalendran, Matthew G. Son, Yuehua Tang, and Yuan Wang · Working paper
Corporate bonds trade in a dealer market with no consolidated tape, so the price of a bond on a given day is an estimate rather than an observation. We build that estimate with machine learning and use it to measure bid-ask spreads, price discovery, and the pricing inefficiencies that survive in OTC markets.
Draft on request
Macho CEOs: Unveiling the Detrimental Impact of Masculinity on Corporate Innovation
Kiyoung Chang, Jongchul Park, Jung Chul Park, Steve Sauerwald, and Matthew G. Son · Working paper
Using deep face-recognition models applied to a comprehensive panel of CEO portraits, we construct a biometric measure of facial masculinity and relate it to firm innovation outcomes. The paper is as much about what image models can and cannot measure about executives as it is about the innovation result.
Draft on request
It is Better to be Yourself: A Network Analysis on Mutual Fund Distinctiveness and Performance
Matthew G. Son · Working paper
I represent the universe of U.S. mutual funds as a holdings network, then measure how common each portfolio is through centrality and how distinctive it is through clusteredness. Funds crowded into commonly held assets underperform by about 60 basis points a year, while funds with localized, distinctive portfolios outperform by about 140.
SSRN
FIN 6776Fall 2026
Learn financial big data workflows and the machine learning toolkit through lectures, labs, and projects you can add to your portfolio.
FIN 4773Not scheduled for Fall 2026
Machine learning concepts in finance, applied to market data through lab projects in R and Quarto.
FIN 4770Not scheduled for Fall 2026
From basic programming concepts to financial analysis in code, using the Unix shell, R, Python, and generative AI.
Notes
on markets, AI, and the tools of the trade
Aug 13, 2026
The tools I've written research code in since 2017 — Python, R, RStudio, VS Code, Positron — and how an outdated cluster finally landed me in Neovim.