Austin Lowey | Machine Learning Engineer

Hi! I'm Austin Lowey,

I'm a Machine Learning Engineer with a Master's in Engineering from Stevens Institute of Technology and over 6 years of engineering experience in Machine Learning, Python, product development, and technical project management leading a cross-functional team.

Projects Contact Me

About

Master's & Bachelor's engineering degrees from Stevens Institute of Technology (3.9 GPA)
⤷ + Carnegie Mellon additional post-graduate coursework in Machine Learning and AI
4 Years of Machine Learning and Software Engineering Experience
⤷ Python, Machine Learning, Computer Vision, Autonomous UAV Swarms, Data Analysis/Visualization
2 Years of Product Development Engineering Experience
⤷ Technical Leadership, Product Development, Project and Risk Management, Stakeholder Requirements

Languages: Python, SQL, HTML/CSS/JavaScript, C++, MATLAB, Bash
Python Libraries: TensorFlow, PyTorch, scikit-learn, pandas, Gymnasium, Matplotlib, Plotly
Databases: PostgreSQL, AWS RDS, dbt, pgAdmin, psycopg2
Mechanical: 3D Printing, 2D & 3D Computer-Aided Design (Creo, Solidworks, AutoCAD), GD&T
Other Technical Skills: Git/GitHub/GitLab, Docker, REST APIs, ROS, Linux

Experience

US Army - DEVCOM

Machine Learning Engineer

Lead a deep reinforcement learning research project to optimize autonomous UAV search speed, translating stakeholder requirements and managing technical strategy and execution
Architected the project’s end-to-end ML application framework, designing a multi-modal neural network and core software components including a custom Gymnasium environment and model evaluation system
Established the MLOps workflow using MLflow and TensorBoard to ensure reproducible experiment tracking
Integrated computer vision models into a R&D UAS software system, adding new target recognition capabilities

July 2023 - Present | Picatinny Arsenal, NJ | Security Clearance: Secret

US Army - DEVCOM

AI Software Engineer

Researched and applied cutting-edge AI and UAS technologies, rapidly prototyping novel solutions
Streamlined data collection and analysis with automated dashboard visualizations
Collaborated with stakeholders to gather and manage requirements, plan program activities, and coordinate tests, ensuring alignment with project goals in an Agile development environment

Nov 2021 - July 2023 | Picatinny Arsenal, NJ | Security Clearance: Secret

US Army - DEVCOM

Product Development Engineer

Deputy program lead and systems integrator for an interservice, $185 million novel munition program
Coordinated technical actions across a 27-engineer IPT & collaborated on system requirements with stakeholders
Accelerated critical-path activity completion by 12 weeks, leading to successful product qualification
Led a root cause analysis project that saved $430,000 and coordinated production/quality for $9.1 million of assets
Coordinated engineering analysis and execution of program’s $3.4 million successful qualification testing series

Dec 2019 - Nov 2021 | Picatinny Arsenal, NJ | Security Clearance: Secret

Edgewell Personal Care

Product Design Engineer (Co-op)

3D printed prototypes and created Creo CAD parts, assemblies, and drawings for consumer products, including brands such as Schick, Edge, Playtex, Banana Boat, Bulldog, and Wet Ones

Aug 2017 - Dec 2017 | Allendale, NJ

Mattel

Product Development Engineer (Co-op)

Developed cost-reduction options for early toy design concepts while managing cross-functional team member requirements and incorporating DFMA engineering design principles to optimize user experience
Designed 3D CAD models of toy parts and mechanisms using SolidWorks and 3D printed design prototypes

Jan 2017 - May 2017 | New York, NY

AKRF

Environmental Engineer (Co-op)

Oversaw environmental drilling to ensure each drilling site complied with safety standards, maintaining a safe and secure work environment, and logged field results using Excel and AutoCAD to support bioswale design

Sep 2015 - Dec 2015 | New York, NY

Featured Personal Projects

Machine Learning - Survivor Winner Prediction (Project Investment: 75+ hours)

• Leveraged machine learning to predict the winner of the reality competition show, Survivor
• Conducted extensive data extraction, transformation, and loading (ETL), including mass web scraping and parsing of the Survivor Wiki, and feature engineering, for data on 700 contestants across 45 seasons
• Integrated OpenAI’s LLM API to conduct automated AI analysis on contestant descriptions, generating social and strategy scores to significantly enhance the machine learning model
• Main Tools Used:
- • Python: scikit-learn, pandas, NumPy, Beautiful Soup/requests, psycopg2, openai (LLM API)
- • Databases: PostgreSQL, pgAdmin
More info on data ETL, feature engineering, ML model training, performance metrics, iteration/refinement, and predictions.

Data ETL & Feature Engineering:

Model Predictions on Test Dataset (showing 25/165 rows -
see GitHub for full dataset):

Data Science - Social Media Trends Tracker (Project Investment: 100+ hours)

• Complete end-to-end, full-stack data application and pipeline.
• Extracts social media posts daily using Python, loads and transforms data using PostgreSQL and dbt, utilizes NLP and Machine Learning to analyze posts/replies/subreddits, and provides visualizations of the trends and insights through a dashboard.
• Main Tools Used:
- • Python, PostgreSQL (initially AWS RDS, then later switched to local solution), dbt, Dagster, NLP/ML
More info on ELT pipeline, mathematical modeling, design choices, etc.

Data Pipeline and Entity Relationship Diagram:

Personalized Spotify Festival Playlist Generator (Project Investment: 200+ hours)

• Enhances Spotify playlist experience with exclusive features not available on Spotify's platform, including better ways to automatically generate large playlists and several new playlist customization options
• Effortlessly create hours-long playlists in seconds using a built-from-scratch, user-friendly interface and dashboard
• Main Tools Used:
- • Back-end: Python (pandas, Beautiful Soup, Spotipy, Plotly, NumPy), REST APIs
- • Front-end: HTML/CSS/JavaScript for summary dashboard (browser), Python PyQt for main GUI (desktop application)
Full, extensive documentation on all available features and usage instructions.
Samples of auto-generated playlists, sorted by music genre.

Demo Video: Demo:

GUI Screens Flow:

Deep Neural Networks with TensorFlow - Multiple Repositories/Projects (Project Investment: 40+ Hours)

• Built deep neural networks using TensorFlow to optimize performance on multiple prediction and classification problems.
• Implemented supervised learning solutions using multiple techniques and architectures, including:
- • Architectures: Convolution Neural Networks (CNNs), Multilayer Perceptrons (MLPs)
- • Hyperparameter Tuning: Keras Tuner, Hyperband, Bayesian Optimization, GridSearchCV
- • Data Augmentation: Spatial Augmentations, SMOTE
- • Regularization: Dropout, Batch Normalization, L2 Ridge Regression
- • Other Techniques: Ensemble Learning, Stratified K-Fold Cross Validation

• Main Tools Used:
- • Python, TensorFlow/Keras, scikit-learn, NumPy, Matplotlib, pandas

• Project Repositories:
1) Predicting Repeat Audiobook Customers for Ad-Targeting (Accuracy: 82.6%)
2) Classifying Images with MNIST Handwritten Digits Dataset (Accuracy: 99.5%)

Reinforcement Learning - Multiple Repositories/Projects (Project Investment: 40+ Hours)

• Developed and trained reinforcement learning agents for multiple problems using Python/Gymnasium.
• Solved the CartPole problem using 3 approaches, all achieving a 100% success rate:
- • Classical Q-Learning with Discretization (NumPy)
- • DQN (Manual Implementation with TensorFlow and NumPy)
- • DQN (API Implementation with Stable Baselines)
• Trained an RL agent for the Atari Breakout game using CNNs to process raw pixel data.
• Implemented a custom RL environment for the Snake game and then trained an agent to successfully play the game.

• Main Tools Used:
- • Python, Gymnasium, Stable Baselines, TensorFlow/Keras, NumPy, Matplotlib

• Project Repositories:
1) Solving the CartPole classic control problem using 3 different approaches.
2) Training agents to play Atari games.

Automated Personal Finance Tools (Project Investment: 30+ hours)

• Automates and simplifies personal finances by projecting future bank and cc balances based on historical and recurring transactions and providing spending analytics and visualizations.
• Main Tools Used:
- • Python (Pandas, Matplotlib, Plotly)
More info on the automation pipeline, auto-generated visualizations, etc.