Feb 2023 - May 2023 | GitHub

● Built an AI system for concise lay summaries of complex biomedical papers, improving accessibility for non-experts

● Engineered a novel approach that extracts dominant sentences and inputs them into fine-tuned PEGASUS-X models

● Achieved better summaries in 60% of the machine metrics such as ROUGE in comparison to the baseline models

Apr 2023 - May 2023 | GitHub

● Programmed a mobile robot using ROS to autonomously find colored boxes and move them to a goal in a specific order

● Utilized Hector SLAM to map the environment as well as localize the robot using only the LiDAR sensor

● Integrated several pretrained computer vision models such as EfficientDet and YOLO to detect and locate the boxes

Oct 2022 - Dec 2022 | GitHub

● Utilized OpenCV to extract digits and symbols from images containing mathematical equations using shape contours

● Trained a baseline CNN using PyTorch on the MNIST and CROHME datasets, resulting in 93.51% test accuracy

● Engineered and tuned custom DCCNNs to classify digits and symbols in math equations correctly 95.47% of the time

Nov 2022 - Dec 2022 | GitHub

● One agent can reach the goal the long way, but pressing a button and opening the door for the other agent is faster

● Programmed Monte Carlo, SARSA, and Q-learning, all of which successfully taught the agents to go through the door

Oct 2020 - Dec 2020

● Used Scikit-learn’s Logistic Regression & bag-of-words to train a stock market predictor model on 500 comments

● Improved the training by developing 2 other models, Logistic Regression with BERT and a neural network with BERT

● Tested the 3 models by cross-validating 6 days of new comments, predicted the stock market will lean mostly neutral