Projects

Graduate Research Student - CERLAB

I delved into the realm of real-time 3D pose estimation, addressing challenges such as occlusion-induced inaccuracies. Drawing inspiration from an ICRA paper, I adapted a diffusion-based approach to explore industry-relevant solutions. Currently, I am pioneering a novel methodology by isolating and training this diffusion-based approach specifically on the H3.6M dataset, with a focus on specialized testing. Through this endeavor, I aim to develop cutting-edge techniques that enhance the accuracy and robustness of real-time 3D pose estimation systems, thereby contributing to advancements in computer vision and related fields. 

Dynamic Multilingual Weather Bot 

In this project, I developed a dynamic weather forecasting application using Streamlit, integrating real-time weather data from the OpenWeatherMap API. Leveraging prompt engineering techniques, I enhanced the application's functionality by generating context for large language models (LLMs) and dynamically requesting responses with placeholders for dynamic data. Additionally, I implemented and evaluated the performance of two different LLMs, TextBison and Gemini-Pro, for processing weather data, providing insights into their accuracy and effectiveness for weather forecasting tasks. 

Network Attack Predictor

In this project, I developed a network attack predictor to enhance cybersecurity measures by predicting and categorizing cyber-attacks. We integrated train and test datasets from the MQTT Kaggle Dataset into a unified Postgres Database table and applied feature engineering principles using PySpark and PyTorch/Tensorflow. We then constructed and deployed ML models, including Linear Regression and Random Forest Classifier, with hyperparameter tuning to boost predictive accuracy. Our network attack predictor demonstrated significant improvements in accurately identifying and categorizing cyber-attacks, thereby enhancing proactive cybersecurity measures. By predicting and categorizing network attacks in real-time, our tool can help organizations preemptively defend against cyber threats, safeguarding sensitive data and infrastructure. 

Actuation of Cardboard Cutouts at Virgini Tech Basketball Stadium

During my tenure as Vice President of the American Society of Mechanical Engineers at Virginia Tech, I spearheaded an initiative to reignite school spirit amidst the challenges posed by the COVID-19 pandemic. Leading a team of seven mechanical engineering students, we embarked on a five-month journey to craft an innovative solution. Leveraging our technical expertise, we meticulously designed and engineered 112 dynamic cutout faces, each equipped with motorized mechanisms to create captivating animations. Through careful planning and execution, we integrated these mechanisms into the cutouts, ensuring smooth and reliable operation. This creative endeavor not only revived the spirit within the student section but also showcased our team's technical and collaborative spirit in overcoming adversity.

New Coverage by Virginia Tech