Grand Teton National Park. Shot on iPhone. 2024. Preston Robinette

Preston Robinette

Hi, there! I am a 5th-year Computer Science PhD student at Vanderbilt University in Dr. Taylor Johnson's Verification and Validation for Intelligent and Trustworthy Autonomy Laboratory (VeriVITAL). I am interested in research related to machine learning applications, reinforcement learning, generative AI, interactive technologies, and security. I am honored to be a recipient of the National Defense Science and Engineering Graduate Fellowship (NDSEG Fellow).
Github     G. Scholar     LinkedIn     CV
preston [dot] k [dot] robinette [at] vanderbilt [dot] edu


Some recent highlights:

Sept. 2024

  • Starting a Student Researcher position at Google on the Gemini Live Team.

May 2024

  • Headed to Google for the summer as a software engineering intern.

Apr. 2024

  • Presented a talk at the International Conference on on Formal Methods in Software Engineering (FormaliSE) 2024 on a Case Study: Neural Network Malware Detection Verification for Feature and Image Datasets.!

Feb. 2024

  • Successfully defended my Qualification Exam on Deep Learning Enabled Methods for Information Hiding!
Show more

Sanitizing Hidden Information with Diffusion Models

Preston K. Robinette, Daniel Moyer, Taylor T. Johnson   •   Code

Case Study: Neural Network Malware Detection Verification for Feature and Image Datasets

Preston K. Robinette, Diego Manzanas Lopez, Serena Serbinowska, Kevin Leach, Taylor T. Johnson   •   Code   •   PDF  •  

Benchmark: Neural Network Malware Classification

Preston K. Robinette, Diego Manzanas Lopez, Taylor T. Johnson
PDF  •   Code

SUDS: Sanitizing Universal and Dependent Steganography

Preston K. Robinette, Hanchen David Wang, Nishan Shehadeh, Daniel Moyer, Taylor T. Johnson
PDF  •   Code

Self-Preserving Genetic Algorithms for Safe Learning in Discrete Action Spaces

Preston K. Robinette, Nathaniel P. Hamilton, Taylor T. Johnson
PDF  •   Code

DEMO: Self-Preserving Genetic Algorithms vs. Safe Reinforcement Learning in Discrete Action Spaces

Preston K. Robinette, Nathaniel P. Hamilton, Taylor T. Johnson
PDF   •   Code

Training Agents to Satisfy Timed and Untimed Signal Temporal Logic Specifications with Reinformcement Learning

Nathaniel P. Hamilton Preston K. Robinette, Taylor T. Johnson
PDF   •   Code

Reinforcement Learning Heuristics for Aerospace Control Systems

Preston K. Robinette, Ben K. Heiner, Umberto Ravaioli, Nathaniel P. Hamilton, Taylor T. Johnson, Kerianne L. Hobbs
PDF

On using real-time reachability for the safety assurance of machine learning controllers

Patrick Musau, et al.
PDF

Automated detection of corrosion in used nuclear fuel dry storage canisters using residual neural networks

Theodore Papamarkou, et al.
PDF

2024 (Fall)

Google   •   (Student Researcher)

2024 (Summer)

Google   •   (Software Engineering Intern)
  • Investigated techniques to enhance instruction-following capabilities in large language models (LLMs) over extended contexts
  • Contributed to the development of datasets, evaluation metrics, and mitigation strategies using prompt engineering in a collaborative repository
  • Designed and executed experiments to benchmark LLM performance and optimize prompting strategies for improved outcomes

2023

Apple   •   Machine Learning Engineering Intern
  • Explored methods to improve data efficiency in Apple’s manufacturing machine learning pipeline used to identify defects in high-resolution manufacturing images
  • Contributed self-supervised learning (SSL) and foundation model knowledge distillation capabilities to a collaborative Git repository, ensuring the usability and repeatability of these contributions across users
  • Conducted experiments utilizing these methods to evaluate detection performance compared to current methodology and established baselines for future work

2022

Department of Defense   •   Cybersecurity Engineer Intern
  • Designed and developed Python scripts to parse and analyze midpoint and endpoint network traffic (PCAPs) using Pandas and regular expressions
  • Created and implemented intrusion detection rules to detect malicious traffic for various common vulnerability exploits (CVEs)
  • Created and evaluated firewall rules to prevent malware attacks on a network
  • Completed various mini-projects related to computer network exploitation, vulnerability research, scanning and exploit development, incident response and data analytics, network forensics, and basic landline and wireless telecommunications networks

2021

Air Force Research Laboratory   •   Reinforcement Learning Intern
  • Investigated the impact of reinforcement learning heuristics on aerospace control systems, an issue arising from the variance of reinforcement learning algorithms
  • Implemented architecture and hyperparameter optimization methods for two aerospace reinforcement learning environments and tasks
  • Improved agent performance (minimum episode length, mean reward, and interaction efficiency) by 200%

2020

NASA Langley Research Center   •   Software Engineering Intern
  • Updated preexisting SAGE III payload software in Python
  • Designed and developed Python scripts to calibrate pre-flight and in-flight telemetry by manipulating and analyzing complex, high-dimensionality data taken from pre-flight laboratory testing and in-flight telemetry
  • Collaborated with data scientists, software engineers, and project managers

2019

NASA South Carolina Space Grant Consortium   •   Undergraduate Research
  • Developed a 3D printed, open source, prosthetic hand controlled via myoelectric sensing and interpretation
  • Designed and implemented control for the hand in C++ by measuring voltages from specific muscles and calculating targeted responses
  • Conducted signal processing in Python to study the relationship between myoelectric signals and individual finger movements
  • Implemented and tested machine learning algorithms to differentiate finger movements with 80% accuracy

2019

Oak Ridge National Laboratory   •   Undergraduate Summer Research
  • Developed a CNN to detect corrosion in spent nuclear fuel canisters with 96% accuracy using PyTorch
  • Analyzed and labeled a large scale dataset of images to be used in training, validation, and testing
  • Created a graphical user interface that highlights corroded sections of uploaded images in a heat map