Projects

This study examines how environmental and geographical factors influence Aphelinus certus parasitization rates. Using Bayesian hierarchical models on USDA data, we found that longer delays between sample collection and experimentation reduce success. Geographical variations also affect reproductive success, underscoring the need for localized pest management strategies.

New York Taxi Co. leverages advanced analytics and predictive modeling to optimize fleet management, addressing challenges in demand forecasting and fare prediction. The project employed ARIMA, SARIMA, and Prophet models for demand and machine learning for fare predictions, aiming to enhance efficiency and customer satisfaction.

Utilized Flask and JavaScript to develop a browser extension that generates customized resume PDFs via LaTeX, tailored to job descriptions on sites like LinkedIn using the ChatGPT API. Set up a server with user authentication for access. Currently exploring Kubernetes to expand availability to the entire MSU campus community

The purpose of this project is to enhance design space exploration algorithms by developing an adaptive AI that reliably identifies 'good' and 'bad' regions in the design space.

Our project stems from the challenges associated with brownfield redevelopment for solar energy, providing a solution to navigate complexities. Recent financial changes, including PILT adjustments, create a conducive policy environment. Aligned with the MI Healthy Climate Plan, our interactive allows users to assign weights to critical variables, offering a dynamic map highlighting optimal sites.

Electric vehicles (EVs) are becoming increasingly popular as people seek to reduce their carbon footprint and save money on fuel costs. In Washington DC, the adoption of EVs has been steadily increasing, and this project aims to predict how this trend will continue in the future.

The project analyzes a dataset containing news articles and highlights for text summarization. It explores data preprocessing steps like removing stopwords and lemmatization. It then evaluates summarization models like Gensim's TextRank algorithm and the T5 transformer from Hugging Face, including details on training and hyperparameter tuning the T5 model.

This project analyzes weather patterns in East Lansing, from 1979 to 2022 using data visualization and machine learning techniques. It explores correlations between weather variables, performs regression and classification analyses, and aims to understand the impact of human population on climate change

With the increased use of OTT platforms (Netflix, Prime, Hulu, etc.), people have a wider range of choices for movies. Ratings and reviews are important factors for the user to decide whether they should watch it or not. Even if a movie has a high rating, it is likely that few people will dislike it. Users’ tastes in movies can be different. So we need some recommendation system which recommends movies of their choice.

Deployed a Deep Learning model on a Raspberry Pi to process live video for monitoring social distancing and mask-wearing. Devised a strategy to auto-install software scripts on Edge Devices via AWS S3, reducing installation time by 25 minutes. Prototypes using Edge Impulse and TensorFlow Lite showed 90% accuracy.

This paper introduces a user-centric approach to optimize encryption, storage, and retrieval in medical imaging. By selectively encrypting sensitive regions of 3D volumes, it reduces computational overhead and enhances retrieval speed by 47% compared to full encryption. Henon maps ensure secure, lossless data retrieval, demonstrated on the MICCAI BrainS18 dataset.