Texas A&M Industrial Engineering | Statistics + Math
Dylan Bago
I build research and data systems across stochastic modeling, NLP, market sentiment, and quantitative finance. I am especially interested in uncertainty, prediction markets, and the practical use of agentic tools.
Brief Summary
Research-oriented builder with a quantitative finance focus.
Academic base
Texas A&M Industrial Engineering student with Statistics and Math minors.
Research focus
Gaussian process simulations and MILP confidence bands for noisy Pareto frontiers.
Technical interests
Operations research, stochastic optimization, NLP, market microstructure, and LLMs.
Currently thinking about
Prediction market volatility and the rise of agentic tools in everyday life.
Fit by Role
Different parts of the same technical profile.
Quant research
Comfortable turning uncertainty into models, simulations, constraints, and measurable coverage targets.
Trading
Interested in prediction markets, real-time updating, fair value, and decision-making under pressure.
Data science
Builds end-to-end pipelines from messy text and market data to benchmarked model outputs.
Relevant Experience
Research and market-facing technical work.
TAMU Operations Research Department
Research Intern | Sep 2025 - Present
- Fit Gaussian process surrogate models to noisy multi-objective function samples.
- Built a PuLP/CBC MILP to find minimal-area confidence bands covering at least 95% of simulated Pareto fronts.
- Automated Monte Carlo trials across noise levels and kernel hyperparameters.
TAMU InfoLab
Research Assistant | Jan 2026 - Present
- Built SpokenCRS, a Python API standardizing conversational recommendation datasets.
- Created CRSDataFrame and TurnWrapper abstractions, reducing onboarding time by an estimated 80-90%.
- Designed unified turn schemas for utterances, entities, ratings, and multimodal metadata.
Aggie Investment Club Quant Division
Social Sentiment Algorithm Developer | Jan 2026 - Present
- Scraped finance Reddit posts with PRAW and classified sentiment using FinBERT.
- Merged sentiment with yFinance data for next-day stock direction labels.
- Benchmarked Logistic Regression, Random Forest, Gradient Boosting, SVM, and LightGBM.
Personal Projects
Selected projects and competition builds.
Pareto Confidence Bands
GP posterior simulations plus MILP coverage constraints for noisy multi-objective optimization.
- Targeted at least 95% simulated frontier coverage.
- Stress-tested across noise levels and kernel hyperparameters.
SpokenCRS
Unified ReDial and Inspired conversational recommendation datasets into one benchmark-ready API.
- Reduced onboarding time by an estimated 80-90%.
- Schema supports 3+ CRS model architectures.
Reddit to Equity Direction
Finance-submitted text, FinBERT sentiment, yFinance labels, and classifier comparisons.
- Scraped 7 finance subreddits across TSLA, AAPL, and AMZN.
- Benchmarked 5 classifiers with GridSearchCV tuning.
Blind Karaoke
HackMIT project using Spotify playback, Whisper transcription, and lyric scoring.
- Built at HackMIT, selected from about 1,000 students internationally.
- Used WER/F1-style scoring for lyric accuracy.
AggieQuant Highlight
Building Texas A&M's quantitative finance community.
I founded AggieQuant to train technical students for quantitative trading and research through probability, statistics, market microstructure, competitions, workshops, and industry engagement.
Skills
Technical toolbox.
Python
Java
SQL
R
C++
PyTorch
TensorFlow
Pandas
PuLP
NLP
LLMs
Git
Bloomberg Terminal
LaTeX
Contact