Garrett S.
engineeringforwarding@gmail.com | github.com/GartcartAbridged public version. Employer-restricted details and embargoed thesis results are withheld, as well as specific dates. Full resume available on request.
Education
Auburn University, Samuel Ginn College of Engineering — Auburn, AL
Auburn University, Samuel Ginn College of Engineering — Auburn, AL
Experience
Graduate Research Assistant — Auburn University
- Built and fine-tuned a YOLO-based computer vision pipeline processing 500+ hours of streaming interstate traffic camera footage for vehicle detection, speed estimation, and seat belt classification, exceeding 0.90 mAP@0.5 on vehicle detection.
- Engineered end-to-end video and image processing workflows in Python (OpenCV, PyTorch, TensorFlow) covering frame extraction, dataset curation, and annotation tooling, reducing dataset preparation time by 28%.
- Designed evaluation frameworks benchmarking 8 model configurations across precision, recall, and inference speed to select production candidates for state transportation smart-corridor deliverables.
- Presented model performance, latency tradeoffs, and deployment recommendations to faculty and external agency stakeholders across a 15-month engagement.
- Containerized training and inference environments with Docker and ran multi-GPU model sweeps on CUDA-enabled hardware, standardizing reproducible runs.
Software Developer Intern — Defense Industry Contractor
- Led a team of 3 interns building Python tooling for serial communication with restricted-access hardware systems, delivering an autonomous test harness still in use as of 2026.
- Developed and tested C++, Java, and Python programs supporting an internal software framework used by 20 engineers.
- Designed GUIs and automation scripts for internal tooling and modernized local file-sharing across company networks, cutting manual test steps by 95%.
Projects
Hybrid Imitation and Reinforcement Learning Driving Agent (MS Thesis)
- Built an end-to-end autonomous driving perception and control pipeline pairing behavioral cloning with a residual reinforcement learning agent, trained on human demonstration laps.
- Authored custom track geometries in Unreal Engine 5 with the Learning Agents plugin, using simulation and synthetic data generation to train and validate policies at scale.
- Designed an automated evaluation harness and quantified results with Firth-penalized logistic regression, Fisher exact tests, and Wilson confidence intervals.
- Full methodology and results available on request once the publication embargo lifts.
ML Benchmarking Pipeline (Wine Quality)
- Built a reproducible benchmarking pipeline comparing 6 ML algorithms under k-fold cross-validation; Random Forest achieved 95% accuracy, with statistical analysis identifying the most predictive features.
Leadership
Founder / President, Technical Interview Association at Auburn University
- Founded and grew the organization to 50+ active members with official ACM financial sponsorship; built a curriculum covering data structures, algorithms, and behavioral interviewing.
- Ran approximately 20 mock technical interviews per semester; in 2025, 80% of active members reported internship or full-time offers at major technology and financial firms.
Skills
- Languages:
- Python, C, C++, Java, SQL, Bash
- ML / Computer Vision:
- PyTorch, TensorFlow, OpenCV, YOLO, scikit-learn, pandas, NumPy
- Simulation / Synthetic Data:
- Unreal Engine 5, Learning Agents Plugin, custom environment authoring
- Tools & Infrastructure:
- Git, Docker, AWS, Linux, CUDA, Ultralytics