Charles Rambo Email & Phone Number

Boston, Ma, Us

Century City

Constructed a systematic convertible bond trading strategy. It considered convertible bond prices, ML-based equity return estimates, security liquidity, and firm credit. Achieved a paper traded return of 11.54% and a Sharpe ratio of 0.84 in 2023. Ordered by returns, the strategy would be placed in third of 17 according to Bloomberg’s FPC ⟨GO⟩ applied to open-ended United States convertible funds.Found equity return estimates using traditional quantitative signals, fractional differenced log-price date, entropy, elastic net, PCA, mutual information, López de Prado’s triple barrier method, and Geurts’ “Extremely randomized trees”.Created a pricing model using the finite difference method. Used Monte Carlo methods to estimate the effect of the 20-of-30 soft call provision contained within most convertible bonds. Experimented with deep learning pricing models using TensorFlow.Built a credit model to estimate "true" letter rating and CDS spread. Gradient boosting techniques are applied to securities within each cluster, where the clusters are constructed using López de Prado's ONC algorithm. In the case of letter ratings, a Kalman filter is used to smooth the outputs. Relative to NRSRO ratings, the estimated credit ratings have a mean absolute error of 1.2 notches post-implementation.Used the above results to construct portfolios using first-order partial derivatives, the underlying equity data, and mean-variance optimization. The covariance matrix is calculated via industry-level factors, Ledoit-Wolf shrinkage, and covariance matrix clipping techniques.Automated the team's weekly reports using Python. Reduced man-hours by approximately 15 hours per week. The team ended up requiring one fewer member as a result.Some experience with TensorFlow. Used for NLP of news reports and for convertible bond pricing.

San Francisco Bay Area

The equity research project examined how categorization bias of firms can result in temporary equity price distortions. Place firms into fundamental groups based on the cosine similarity between business descriptions within firms’ 10K filings. Then compare each firm's group return with its GICS industry return. If the fundamental group outperformed the industry, long the firm's equity. In contrast, if the fundamental group underperformed the industry, short the firm's equity. This strategy achieved an out-of-sample return of 4% and a Sharpe ratio of 0.70 within the period 2019-2020.This project was a partial replication and extension of the paper "Categorization Bias in the Stock Market" by Krüger, Landier, and Thesmar. The alternative method to gauge similarity was based on data obtained from Hoberg. Analysis conducted in Python using MSCI US equity data. Extension of applied finance project.

San Francisco, California, United States

Used mathematical and statistical modeling in Python, including Monte Carlo simulations, to calculate risk scores for risk events, like wildfires and explosions, caused by the firm’s business operations.Conducted the preliminary analysis for the use of the Pareto distribution for consequence modeling and wrote the class for the distribution.

Escondido

Tutored statistics, calculus, linear algebra, GRE mathematics subject test, and Society of Actuaries’ Exam P.Wrote two self-published mathematics books, which were designed to help students study for the GRE math subject test. Created www.rambotutoring.com, which contains hundreds of pages of original and free mathematics content.

Master'S Degree, Financial Engineering

Received MFE Award for Academic Excellence

Bachelor Of Arts, Mathematics

Education record