M.Sc. in Machine Learning and Signal Processing with experience in the application and development of deep reinforcement learning, statistical signal processing, and machine learning algorithms to solve complex problems. I am excited to continue learning and applying these techniques to new domains.
Delivered phase I SBIR contract for autonomous robotic radiation surveying. This utilized a graph transformer model and reinforcement learning to enable LAMP system navigation and source seeking in unknown environments
•
Led full-stack development of detector API, from low-level C++ readout firmware to React-based operator interface
•
Led design and launch of nuclear power plant data management platform, automating compliance reporting workflows
•
Architected ROS-based sensor package that integrates data from multiple dosimeter types, enhancing LAMP system's radiation measurement accuracy and expanding addressable market segments
•
Serve as technical lead for flagship commercial product, debugging production issues, and implementing 10 critical feature enhancements
•
Engineered automated calibration tool for detector systems that reduced pre-shipment QA time by 5 hrs/system ensuring measurement accuracy across deployed units
Thesis: Proximal Policy Optimization for Radiation Source Search
•
First author on paper, in collaboration with another research university
•
Constructed novel deep reinforcement learning architecture in Python / Pytorch that achieved a success rate of 95% in a complex nuclear source search task outperforming gradient search by 68%
•
Built OpenAI Gym radiation simulation environment that used a visibility graph for shortest path calculation with polygonal obstructions
•
Implemented information-theoretic control algorithm from the literature that utilized a Bayesian particle filter for the radiation source search comparison.
•
Mentored 3 NSF-funded undergraduate students on computational modeling research projects and ran lab meetings for 15 students
•
Presented research results and project progress at 2 annual reviews for funding agency
Designed instrument life cycle test protocol in MATLAB for main operational unit that revealed mechanical design flaw resulting in component redesign that increased instrument lifespan by 9%
•
Soldered and tested printed circuit boards used in the robot control system
Created data cleaning script in Python using Pandas to remove duplicates and flag feature input errors, used in an exploratory data analysis to assess efficacy of potential company asset
•
Presented investigative report of company asset performance to management leading to integration of asset into product pipeline
•
Proposed 2 novel data sources for use in the predictive modeling
