Franklin Wohnoutka

Posted: Fits & Hauls — iOS Social Media App

A full-stack social media app for sharing outfits and fashion hauls, built with React Native (Expo) and shipped to the App Store. Features include a social feed, follow system with public/private accounts, real-time notifications, wardrobe calendar, and cloth tagging system, backed by a Supabase/PostgreSQL database with row-level security and real-time subscriptions. Built and deployed independently, including native builds with EAS.

Swipe — Video Download Web App

A collaborative web app built with Flask and Azure that allows users to download videos and audio from YouTube, Instagram, and TikTok. Features include, Google OAuth authentication, Azure Blob Storage for per-user file management, and a responsive Bootstrap frontend with dark/light mode support. I primarily served as the tooling scripter for the project.

MISSFIT — Magneto-Ionization Spacecraft Shield for Interplanetary Travel

An interdisciplinary student-led research collaboration at Drake University designing a magnetic shielding system to protect spacecraft crews from solar wind and galactic cosmic ray radiation. I led the computational subgroup during the spring of 2026, contributing to relativistic particle tracking and graphing code written in C and MATLAB. View our research poster.

Random Search for Treasure on Connected Graphs

A mathematics capstone project inspired by the game of Cops and Robbers. We designed a game where a hunter traverses a graph randomly until finding buried treasure, then computed the expected capture time for every possible starting position. Surprisingly, our results showed that hunter starting position has no effect on average expected capture time. View our research poster.

Modeling of Large-Scale Deportation in Urban Areas

An Agent Based Model built in NetLogo investigating how large-scale immigration enforcement impacts fear and crime in urban communities. Results showed enforcement does not significantly reduce crime rates but increases community fear and widens disparities between targeted and non-targeted populations. View my research poster.

Bird Species Image Classifier

Fine-tuned three CNN architectures, AlexNet, ResNet, and VGGNet, to identify different species of birds at high accuracy. This type of model could be used in something like a phone app that identifies birds in the wild, or in biological labs to classify specimens. AlexNet performed best overall, getting only 2 predictions wrong.

MLB Batter Performance Predictor

A machine learning model that predicts batter outcomes given pitch characteristics using Statcast data. Built using Random Forest, Gradient Boosting, and XGBoost regressors on pitch physics data pulled via the pybaseball library. Initially trained on Luis Arraez but designed to work for any MLB player.

Cause of College Alcohol Consumption

A machine learning study predicting weekend alcohol consumption among college students using KNN, Decision Tree, and Random Forest classifiers. Conducted a feature importance analysis by systematically removing columns to identify which factors — such as social activity, age, and parental education — most influence drinking behavior.

Java Battle Simulator

A text-based RPG battle simulator built in Java demonstrating core object-oriented programming concepts including inheritance, encapsulation, polymorphism, and interfaces. Players choose a character and weapon then fight through three levels of monsters with stat upgrades between rounds.