Hi, I'm Jonah Engelmann

AI · Data · Tech Consulting. MSE in Data Science at UPenn, focused on using data and AI to solve problems in places where they haven't been used before.

Experience

Jun 2026
–
Aug 2026
Artificial Intelligence Intern
Mortenson

Built a team of AI agents at Mortenson that automates the time-intensive process of putting together project proposals, turning a job that used to eat up an entire work week into one that takes about two hours.
- Built a system of five AI agents that work together to write project proposal resumes automatically, cutting a process that used to take over 30 hours of manual work down to about 2 hours.
- Studied how the proposal process worked step by step and broke it down into pieces an AI agent could actually handle, so the system fit into the way the team already worked.
- Designed a mockup of the finished tool so the team could see and react to how the agents would cover the whole proposal process before any of it was built.
Jun 2025
–
Sep 2025
Data Science Intern
Mortenson

Delivered end-to-end data science solutions at Mortenson, building machine learning pipelines, a Flask app for executives, and Fabric and Power BI analytics to improve labor forecasting and financial decisions.
- Built an adaptive ML pipeline that cleans and engineers features from construction task data, then tests combinations of preprocessing and modeling techniques, from neural networks to gradient boosting to regularized linear models, to beat human-estimator baselines.
- Benchmarked model accuracy against human estimators and packaged the best-performing models for fast, reproducible predictions.
- Built a Flask web app where estimators enter project details to get instant hour and productivity predictions, compare them against historical benchmarks, and browse similar past projects — demoed to Mortenson's President.
- Automated the extraction of financial data from hundreds of scattered documents, representing billions of dollars in projects, into clean spreadsheets and Python-generated charts comparing performance across projects, years, and teams.
- Connected Microsoft Fabric to Power BI to build multi-page dashboards tracking solar and renewables project performance for company executives.
Mar 2024
–
Jun 2025
AI Researcher
The Cloud Lab, SCU Department of Computer Science and Engineering

Led a research project resulting in a peer-reviewed publication demonstrating that LLM coding performance is often driven by memorization rather than reasoning, using ELEET—an automated evaluation framework I built to run large-scale, reproducible benchmarks across modern language models.
- Co-authored a peer-reviewed paper (IJCNLP-AACL 2025) analyzing reasoning claims in large language models through systematic evaluation of code-generation tasks
- Built ELEET, an end-to-end evaluation framework that standardizes coding problems, generates prompts, executes model-produced code, and verifies correctness automatically
- Ran 1,000+ controlled evaluations on LeetCode-style problems, isolating recency effects and prompt rewording sensitivity to distinguish memorization from reasoning
- Scaled benchmarks across open- and closed-source LLMs, reducing manual evaluation time by ~90% while ensuring consistent, reproducible comparisons
Jun 2023
–
Sep 2023
Database Intern
MN Veterans Hospital, Minneapolis Adaptive Design & Engineering

Built a full-stack app for veterans in stand-up wheelchairs that converts accelerometer data into precise angles and positioning/standing visuals, recognized with SCU’s REAL Award.
- Built a full-stack app for the Minneapolis Adaptive Design & Engineering (MADE) program to analyze accelerometer data from stand-up wheelchairs used by veterans.
- Converted raw sensor streams into precise angle calculations and clear visuals of positioning and standing intervals, with upload, visualization, and export tools.
- Implemented a Python/Django backend with REST APIs for storage and retrieval and an HTML/CSS/JavaScript frontend
- Work recognized with SCU’s REAL Award.

Publications

Large Language Models Exhibit Limited Reasoning Ability on Coding Problems

Jo, J., Engelmann, J., & Choi, S. · IJCNLP-AACL · 2025

We show that LLMs' apparent success on coding problems is better explained by memorization than reasoning, using recency effects and prompt perturbations on LeetCode questions.

Featured Projects

Stock Sentiment vs Market Movement

Analyzed how Reddit investor sentiment aligns with stock market behavior using NLP and time series analysis.

MorningMotivationBot

Flask web app that delivers personalized motivational messages, haikus, images, and music based on user mood.

Other Projects

Titanic Data Visualization

Tableau analysis of Kaggle’s Titanic dataset to surface survival patterns by class, gender, and age.

Wine Quality Analysis

Modeled wine quality using seven supervised ML algorithms with cross-validation and grid search.

NBA Player Predictions

Predicted NBA draft likelihood using SCU athletes’ performance and combine metrics.

School Course & Teacher Relational Database

Designed a MySQL schema to manage enrollments, students, professors, departments, and courses.

Ultimate Frisbee Analysis

Statistical analysis pipeline and relational database for multi-year ultimate frisbee data.

iOS Alert App — Beacon

SwiftUI app with Firestore backend for rapid alerts and live location sharing with selected contacts.

Social Network App (C++/Qt)

Desktop social networking prototype with friend suggestions and messaging.

Alzheimer’s Brain Scan Modeling (Wharton DSA)

Cleaned, repaired, and visualized complex brain scan data; modeled dementia indicators.

Baseball Outcome Modeling (UMN ML Summer Program)

Trained six classifiers for baseball prediction with dimensionality reduction and tuning.

Contact

Best way to reach me is by email. I'm currently exploring summer opportunities in tech, AI, and data consulting, and always happy to connect with people working in AI, analytics, or consulting.

Email Me

Hi, I'm Jonah Engelmann

Experience

Publications