Hacking Deep Reinforcement Learning Agents to Fail with no Prior Encounter- Elevate the practicality of policy induction attacks against deep reinforcement learning agents by removing their reliance on environment manipulation or prior access to the victim agent- Integrate the state-of-the-art in adversarial policy training with population-based training, meta-optimization, and model extraction attacks to achieve a time-efficient attack- Demonstrate a successful attack and victory against AlphaStar and AlphaZero, investigating future directions for developing defenses and early detection mechanismsMissile Defense Agency (MDA) - Deep Reinforcement Learning Enabled Warfighter Assistant- Architected a full-stack framework of explainable deep reinforcement learning for missile defense, centered on a new method of neural tree perturbations from SHAP and LIME- Developed prototype agents for command, control, and engagement tasks in regional missile defense, countering the high unpredictability and limited decision times with robust training- Submitted a full Phase I proposal for topic MDA22-T004, with comprehensive strategies and weekly plans for R&D and commercialization for military and private sector customers

Lockheed Martin Advanced Technology Laboratories (ATL) is an applied research and development center for scientific advancement in the areas of autonomy, robotics, artificial intelligence, command & control, human-machine symbiosis, spectrum operations, hypersonics and cyber.Their customers include U.S. government service laboratories and the U.S. armed forces, the Defense Advanced Research Projects Agency (DARPA), and the Intelligence Advanced Research Projects Activity (IARPA) to maintain technology dominance.

- Selected as the first non-prior service reservist in history to conduct Offensive Cyber Operations, with strong support from the Air Force Reserve Command and Headquarters Air Force- Plan and conduct Defensive Cyber Operations to protect cyberspace capabilities, data, networks, net-centric capabilities, and other systems through aggressive internal threat hunting, forensics, reverse engineering, and malware analysis- Research and develop new capabilities, tools, Tactics, Techniques, and Procedures (TTPs) to enhance power projection in, from, and through cyberspace

Omniscient Cryptocurrency Trading with HyperNEAT and DRL- Developed trading agents that can mimic the learning process of human traders identifying market signals from numerous information on the world, at computational speeds and scale- Incorporated Hypercube-based NeuroEvolution of Augmenting Topologies with reinforcement learning to efficiently evolve large neural networks with noisy and nuanced information- Implemented web scraping to allow independent data collection and selection based on the neural connectivity patterns evolved, analyzing them to further investigate market dynamicsGenetic Programming Framework for Symbolic Regression Methods- Developed and maintained a genetic programming framework to interface with the Eureqa algorithm, Bayesian machine scientist, PySR, and AI Feynman, allowing experimentation and synthesis to support the development of new symbolic regression methods- Implemented measures to address common challenges such as algebraic equivalence, parsimony-accuracy tradeoff, quality diversity, and degree of compression

- Researched and developed a new multi-agent reinforcement learning framework to scale up and extend the winning AI pilot of the AlphaDogfight Trials, to maintain performance while controlling multiple, heterogeneous collaborative combat aircraft- Trained prototype agents for 2-on-2 aerial dogfights by shaping rewards for both combat performance and flight safety, initiating the work to induce complex collaborative behavior- Supported Shield AI’s success in proceeding to Phase 2 of the program in a down-select from eight to three contractors, cutting Lockheed Martin and Boeing

- Worked to develop the first viable electrohydrodynamic (EHD) propulsion aircraft: a plane with no moving parts, using the movement of ions from the air- Struggled to find components, equipment, and the right expertise (handling high voltage electricity at 40kV) as I was in Vietnam with a significant language barrier- Managed to build a test thruster from motorcycle parts and collaborate with a local Australian university to safely conduct my experiments