Implemented iterative and gradient descent phase reconstruction algorithms to generate multi-layer 3D tweezer arrays using a phase spatial light modulator.

Training a U-Net encoder-decoder phase reconstruction neural network to generate phase maps for 3D tweezer arrays.

Implemented a pipeline to characterize 3D trap arrays in an experimental setup.

Designed transformer and structured selective state space models to decode the rotated planar code, a type of quantum error correction code.

Implemented and trained the models using PyTorch to decode low distance rotated planar codes.

Scaled decoders to higher distance codes using state compression techniques.

Trained a federated QCNN using the Qiskit Machine Learning library, achieving 89% simulator test accuracy and 70% IBM QPU test accuracy on the MNIST dataset.

Implemented differential privacy to obfuscate sensitive client data, and performed a hyperparameter search to find an appropriate level of privacy.

Contributed to a white paper detailing relevant literature and proposed approaches on the weapon target assignment problem, which resulted in the team's acceptance to the competition.

Played a leading role in brainstorming and implementing approaches for automatic scheduling and coordination of advanced weapon systems.

Architected, implemented, and trained several approaches to reduce damage to high value assets, including a Deep Q-Learning agent and heuristic-driven Greedy agent.

The W&M team won 3rd place and $20,000 in prize money.

Approximated arbitrary quantum operators using the Pauli product rotations, exponentiated elements of the Pauli group.

Transformed problem into nonconvex positive semidefinite programming problem, and optimized using a trust-region approach.