This portfolio contains my work across various domains such as robotics, machine learning, deep learning, computer vision, natural language processing, data visualization, optimization & statistical analysis, etc.
During my R&D internship at New Bedford Research & Robotics, I developed and deployed robotic solutions, including automated pick-and-place systems, object position prediction, and custom 3D-printed components. One of the highlights was using the robot in a ribbon-cutting ceremony event, which was covered in local news.
In this project we visualize mass shootings in USA for the years 2014-2022.
This project utilizes an end-to-end Retrieval Augmented Generation (RAG) model powered by LangChain, alongside traditional NLP techniques like TextRank, SimHash, and MinHash to summarize and analyze literary texts. Results demonstrate that RAG outperforms traditional methods in both summarization and question answering tasks.
This project highlights the effectiveness of parallelism in handling computationally demanding tasks by using techniques like OpenMP, MPI, and MPI broadcast, combined with YOLOv3 for human face detection in videos. Results show that MPI broadcast achieved the best optimization, with the highest performance observed using 8 processes.
This project uses advanced text classification techniques, combining statistical methods like TF-IDF with Logistic Regression and Naive Bayes, alongside deep learning models such as CNNs, LSTMs, and Transformers. The results show that deep learning models significantly outperform traditional methods in classifying consumer complaints.
This project is a calorie tracking app that allows users to log meals, track calorie intake, and receive health feedback. It features CRUD (Create, Read, Update, Delete) operations for meal management, user registration, and integrates with a food database. Built with Python, Kivy, and MongoDB, the app helps users monitor their health over time.
This project analyzes gold price trends using historical data and applied SARIMAX and LSTM models for forecasting. Results highlighted SARIMAX is better accuracy in capturing patterns, revealing factors like seasonality, demand, and mining challenges impacting gold prices.
This project experiments with derivative-free optimization methods in MATLAB to estimate VFA T1 mapping in MRI. Results demonstrate that Novifast outperforms traditional techniques in both accuracy and speed, providing efficient T1 estimation even under varying signal-to-noise ratios (SNRs).