Project: “Recorded recurrent deep reinforcement learning guidance laws for intercepting endoatmospheric maneuvering missiles” (Qiu et al., 2023)- Implemented novel reinforcement learning algorithm, RRTD3, from the given research paper.- Built the 2-DOF missile interception simulation environment, as outlined in the paper, using the Gymnasium library in Python.- Created the architecture for the RRTD3 agent using PyTorch, as well as a Sequential Replay Buffer to store its simulated experiences.- Achieved better performance on the interception task than RLLib’s base TD3 agent by over 30%.

Project: Online Multiplayer Game Theory Research for Mind, Machine, and Mathematics Laboratory - Designed a pairing scheme for participants between rounds in the experiment - Researched psychological effects of cooperation and competition on group dynamics

AIOps: Event Correlation and Noise Reduction- Wrote a Python script to restructure and extract relevant features from a large dataset of IT log incidents in a Google Colab environment, which significantly reduced training time and improved accuracy for the model.- Built two unsupervised machine learning models using the TensorFlow library that took in the cleaned dataset as input, identified and eliminated noisy data, then clustered events of the same type with ~77% accuracy. - Integrated my models into the AIOps team pipeline, increasing the degree of automation for IT operations.

Project: Deep Learning for Rodent Behavioral Analysis- Experimented with a deep learning tool called B-SOiD in order to analyze video and spatial data of rats performing a skilled reach with the goal of gaining insight on the effects of dopamine on motor control.- Trained B-SOiD to create graphical representations and video snippets of the rats’ sub-movements during the skilled reach, which helped the team identify the most crucial neural pathways for performing such movements.- Improved analysis speed, reproducibility, and experimental throughput for the research team by using the B-SOiD workflow pipeline I established over the course of my second semester with them.