Projects
An FPGA-based JPEG image compression system using DCT and UART communication, built as part of Digital Systems course.
A neural network-based text generator trained to predict the next character using previous k characters
Image reconstruction pipeline using Random Fourier Features and matrix factorization for compression and upscaling with over 32 PSNR.
Used TSFEL and Decision Trees on UCI HAR dataset for activity classification, also explored LLM-based zero/few-shot recognition techniques.
A Flask-based web app to analyze electoral bonds data via MySQL backend, PDF-to-CSV conversion, and dynamic charts and tables.