Marshall Ratliff

Originally from Montclair, NJ, I completed my undergraduate studies at Duke University in May 2015, receiving a B.S. in Mathematics with distinction and a minor in Finance. While at Duke, I worked as an undergraduate researcher at the Information Initiative at Duke (iiD) studying Topological Data Analysis (TDA), machine learning, classification tasks, and exploratory data analysis focusing predominantly on Music Information Retrieval (MIR). Using sliding window embeddings to map songs to curves in high-dimensional feature spaces, I applied various topological and geometric analysis techniques alongside machine learning methods to classify songs by genres and generalized labels, identify beat structure and segmentation of songs, identify cover songs, and more generally to visualize songs as curves in 3D, leading to the online application loopditty.net. In addition, using a hierarchical clustering data structure known as the cover tree, I developed a scheme for hierarchical labeling for understanding the divergence and similarities of genres and subgenres of music with respect to relevant musical and qualitative features. While at Duke, I also worked part-time as a Data Analyst for cold-storage and transportation logistics provider Lineage Logistics on data-based optimization problems including transportation, freezer cell operation and utilization, and storage regimes and methodology.

After graduating I went to work as an electronic trader for a major global liquidity provider and high-frequency trading firm, Flow Traders, focused on market making in exchange-traded funds (ETFs). Flow Traders has been recognized as the #1 ETF Market Maker in Europe and Asia for a decade and has steadily increased its presence in the US, being recognized as the #2 ETF Market Maker in the US for the last 5 years. My personal focus was on international developed equity derivatives where I was registered market maker in hundreds of US-listed ETFs, often as a top 3 participant in terms of volume traded and creation/redemption activity, comprising over 100,000 individual companies internationally and domestically. At Flow Traders, my job was to maintain passive buy/sell quotes and active trading strategies in all names in which we made markets, develop pricing models for new products and improve pricing of current products, and work on new trading strategies for the desk and company. Given that a large number of companies I traded were closed, a significant amount of statistical modeling and back-testing was necessarily part of the job to determine the fair value of equities outside of local trading hours.

In the MIDS program at the iSchool at UC Berkeley, I am expanding on my background in time-series and exploratory data analysis and machine learning to develop an arsenal of approaches to data-based problems in the financial, industrial or technological sectors. I am working on expanding my past work in music data analysis to further understand the evolution of musical styles and genres over time and through specific artists and songs. I am also excited about building more robust financial models that take advantage of cutting edge techniques and big data to improve profitability and decrease risk as well as how to use financial transaction data to identify potential systemic market risks for trading opportunities and for regulators. Finally, I am working on a new project of understanding how gene expression is related to climate, nutrients and other factors (“the phenome project”) to develop more efficient farming methods and more nutritious crops. I am interested in problems of portfolio risk management, improving portfolio alpha, financial regulation, music classification, agronomy, and efficient urban farming.