An end-to-end Retrieval-Augmented Generation (RAG) web app built to interactively answer questions related to my peer-reviewed publications. This project is an AI research assistant, which retrieves relevant responses from dense academic papers, researching Supermassive Black Holes, Active Galactic Nuclei (AGNs), Blazars, and QPO analysis of blazars.
S&P 500 represents the 500 most valuable companies of the US stock market. In this project, I utilize multiple statistical and machine learning models to forecast the market trend. See for yourself how different models stack up against eachother and how effective are they to forecast impending market crash.
The raw data from the Fermi-LAT telescope is analyzed for classification of BLL and FSRQ types of blazars. Three classifier ML algorithms were trained for it, Decision Tree (DT), XGBoost DT (GBDT), and Random Forest (RF). The GBDT classifier was found to be the most accurate with accuracy >90%.
Transfer the artistic style of a 'style' image to a 'content' image. In this project, I utilize a pretrained image layer filtering algorithm VGG19 to transform the style of an image.
This project implements a N-Gram model to learn the word pattern from a dictionary to solve the Hangman Word Game. The N-Gram model is currently ~66% accurate when tested with training dictionary.
In time-series astronomy, periodicity study is one of the most useful tools to understand an astrophysical system. Research papers on periodicity study often show phase-folded plots to emphasize the presence of periodicity. Depending on the quality of the data points, we can use the phase-folded light curves (LCs) to further enhance the accuracy of the observed period. I discuss an amplitude minimization of the phase-folded LC, which can enhance the accuracy of observed periods in astrophysical LCs.
In Prep...