Ben DeJong

Professional Summary

Computer Scientist and Mathematician with hands-on research experience in AI, LLMs, HPC, and compression. 5+ years building performant software in Python, C++ (C++23), and Rust, plus 3 years working with machine learning frameworks. Adept at using statistical and mathematical methods to solve complex, real-world problems. Clear communicator with a track record of cross-functional collaboration and technical tutoring. ICPC, Putnam, and NCAA calculus competitor.

Education

Elmhurst University — 2021–Present (Anticipated Graduation: May 26, 2026)

B.S. in Computer Science and Mathematics (Double Major)
GPA: 3.845/4.00
Mathematics and Programming competition teams
7 times Dean's List
Transfer Excellence Scholarship
Honors Program Member
ICPC, Putnam, and ACCA calculus competition competitor.

Technical Skills

Machine Learning

Pytorch
Scikit-Learn
NumPy
PANDAS
Backtesting
Model Evaluation
Gen AI

Mathematics

Calculus I–III
Probability
Mathematical Modeling
Real Analysis
Differential Equations
Discrete Math
Data Visualization
Numerical Methods

Programming

C++ (C++23)
Python
Rust
CUDA
Matplotlib
Seaborn
Git
Docker
Gen AI

Languages

English
Spanish

Work Experience

Tutor – Elmhurst University (7 months) Aug 2025 – Present

Explained Calculus I-III, statistics, and algebra rigorously and intuitively to 40+ students.
Managed tutoring schedules and maintained detailed student progress reports.

Researcher – Argonne National Lab (4 months) May 2025 – Aug 2025

Benchmarked compressing collective communication libraries on HPC systems across 128 multi-GPU compute nodes.
Built Python data pipelines to parse outputs, analyze performance, and visualize scaling behavior, creating insights.
Identified and explained unexpected behavior through empirical scaling tests, targeting problems.
Fixed flawed benchmark code by modifying C++/CUDA timing code, enabling accurate millisecond-level measurements.
Presented and defended findings at two poster sessions.

LLM Researcher – Elmhurst University (7 months) May 2024 – Dec 2024

Scraped and cleaned datasets to fine-tune LLMs, improving accuracy and reliability.
Designed and evaluated prompts to improve usefulness, reliability, and mitigate prompt-injection attacks.
Collaborated with a cross-functional team to give an accurate, documented, reliable product to an external company.

Tech Support & Sales – Perry’s Music (6 years) Nov 2018 – May 2025

Assisted 1000s customers, negotiated 100s of sales online and in person in written and spoken English and Spanish.
Configured devices and solved technological computer problems.
Managed a 200 item inventory using Microsoft Excel and created online item listings, enabling targeted advertisement.

Projects

Custom Compiler and Programming Language

Developed a formal grammar for a statically-typed imperative language.
Classified identifiers, numeric literals, operators, and reserved keywords using a custom lexer implemented with regex.
Validated syntax and expressions using a custom recursive-descent parser operating on a token stream.
Applied compiler front-end techniques including token lookahead, grammar-driven parsing, and error detection.

Market Microstructure and Equilibrium Modeling

Modeled realistic market matching dynamics by building a limit order book simulator with priority queues in Python.
Analyzed emergent price formation through large-scale stochastic trade simulations.
Learned market equilibrium prices with 95% accuracy using a neural regression model (PyTorch) trained on simulated markets.
Predicted buyer-seller interactions with a 94% accuracy using nonlinear differential equations solved using RK45 and Euler methods.

Car Price Predictor

Built an end-to-end data pipeline, including web scraping, cleaning, and preprocessing with BS4 and Pandas.
Performed feature engineering by tokenizing variables for supervised regression.
Achieved 93% accuracy utilizing scikit-learn decision trees.
Compared PyTorch neural networks against classical models to analyze tradeoffs and complexity.

Blackjack Model

Applied Deep Q-Learning and classical algorithms to learn near-optimal policies, achieving a 45% win rate.
Implemented a Blackjack environment from scratch as a Markov Decision Process in Python.
Evaluated policy performance through large-scale simulation and statistical analysis.

Keywords

Computer Science, Software Development, Machine Learning, Artificial Intelligence, LLMs, High-Performance Computing, Data Analysis, Python, C++, Rust, CUDA, Pytorch, Scikit-Learn, NumPy, Pandas, Mathematical Modeling, Statistics, Calculus, Differential Equations, Data Visualization, Git, Docker.