Filter Results:
(select to filter results by technology)
Projects:
(click on Photo Gallery for pictures)
Trending Stock Market WebSite
Northeastern University
JavaScript
React
Express
Axios
MongoDB
Node
MERN
HTML
CSS
Bootstrap
JWT tokens
Redis
git
linux
Photo Gallery
- Architected and developed a full-stack stock market trend analysis platform using the MERN stack, integrating real-time financial data display with interactive features such as commenting, liking, and viewing posts, leveraging technologies like MongoDB, Express.js, React.js, Node.js, and Bootstrap for a responsive design.
- Implemented advanced web functionalities including JWT-based state and session management, Redis caching for optimized data retrieval, OAuth for secure third-party authentication, and data encryption, ensuring a secure and seamless end-user experience with efficient load times and robust security protocols.
- Engineered a dynamic data visualization tool in Python utilizing libraries such as Pandas for data manipulation, Matplotlib for plotting, and NumPy for numerical computations, incorporating Tkinter to create an interactive GUI that enables users to select datasets, graph types, and variables for customized visual analysis.
- Designed and implemented a robust graphical user interface using Tkinter, allowing for intuitive user interaction with complex datasets, and integrated Matplotlib to render a variety of plots, such as scatter, bar, and line graphs, facilitating insightful data exploration and pattern recognition.
- Developed an Android application called "Smart Saver" that enables parents to monitor their child's chores and rewards.
- Incorporated sensors and features such as GPS, camera, intents, notifications, recycler views, and fragments to enhance the UI.
Portfolio
Northeastern University
JavaScript
React
Next.js
Node
HTML
CSS
Tailwind
JWT tokens
SSL/TLS cert
git
linux
keyframes
Photo Gallery
- Architected and developed a full-stack stock market trend analysis platform using the MERN stack, integrating real-time financial data display with interactive features such as commenting, liking, and viewing posts, leveraging technologies like MongoDB, Express.js, React.js, Node.js, and Bootstrap for a responsive design.
- Implemented advanced web functionalities including JWT-based state and session management, Redis caching for optimized data retrieval, OAuth for secure third-party authentication, and data encryption, ensuring a secure and seamless end-user experience with efficient load times and robust security protocols.
- Developed an intuitive Android weather application using Java and Kotlin in Android Studio, integrating Retrofit for robust network operations, parsing JSON data from a third-party API, resulting in a seamless user experience for real-time weather updates.
- Implemented a RecyclerView to efficiently display daily weather forecasts, incorporating Material Design principles for a visually appealing interface; optimized app performance through meticulous testing and debugging, ensuring reliability and high user satisfaction.
Tuiter Website (Twitter)
Northeastern University
JavaScript
React
Express
Jest
Axios
MongoDB
Node
MERN
HTML
CSS
Bootstrap
Photo Gallery
- Developed a social networking website, leveraging JavaScript, TypeScript, Bootstrap, React, and Mongo DB on a Node.js server, incorporating RESTful Webservices to serve HTTP requests, and enriched with features like boards, likes, tweets, and follow options.
- Implemented user sessions and achieved 80% code coverage using Jest and React Testing Library for enhanced quality assurance.
- Constructed an adventure game using MVC principles and object-oriented design patterns to make the gameplay realistic.
- Tested project modules using a JUnit 4 test suite, achieving 86% coverage. Implemented the game's view using Java Swing library.
Credit Card Default Prediction
Northeastern University
Python
Matplotlib
Numpy
Pandas
Scikit-Learn
PyTorch
TensorFlow
machine learning
- Utilized logistic regression, decision trees, random forests, and neural networks in Python for credit default prediction. Employed PCA and Elastic Net for dimensionality reduction, stratified K-fold for model evaluation, and LIME for model interpretability.
- Conducted in-depth analysis on credit default factors, utilizing confusion matrices and ROC curves for effective model evaluation and performance optimization. Enhanced accuracy through oversampling in unbalanced classes using Adsyn and weighted models.
Help Robots Navigate
Northeastern University
Python
Matplotlib
Numpy
Pandas
Scikit-Learn
PyTorch
TensorFlow
machine learning
- Advanced Machine Learning in Robotics: Led the creation of a predictive model with Bidirectional LSTM, K-Nearest Neighbors, and Random Forest, utilizing IMU data for floor surface recognition in robotics.
- Data Analysis and Preprocessing Expertise: Managed complex multivariate time series data with normalization and sequence splitting, enhancing robotic navigation and safety through rigorous accuracy, precision, recall, and F1-score evaluations.
Image Generator using Transformers
Northeastern University
Python
Scikit-Learn
PyTorch
TensorFlow
machine learning
Transformers
- Diffusion Model Implementation for Text and Image Generation: Executed advanced machine learning techniques to set up diffusion models, focusing on text-to-image and image-to-image generation. Utilized the diffusers library from Hugging Face and pretrained models for image creation, demonstrating a deep understanding of diffusion probabilistic models as discussed in Lecture 11 of Week 7.
- Leveraging Hugging Face and Google Colab: Employed Hugging Face's diffusers library and Google Colab's GPU resources to efficiently run complex diffusion models. Engaged with Hugging Face tutorials to deepen knowledge in automated pipelines and custom implementations, showcasing skills in adapting and applying cutting-edge AI technologies.
Image Classification using CNN
Northeastern University
Python
Scikit-Learn
PyTorch
TensorFlow
machine learning
CNN
- Advanced Image Classification Project: Successfully implemented and compared neural network architectures for image classification on MNIST and CIFAR datasets, including fully connected models (9,446,402 parameters), CNNs (6,026 parameters), and a pretrained ResNet18 model (11,177,538 parameters). Demonstrated expertise in model training, achieving an accuracy of 88.8% overall, with class-specific accuracies up to 92.2% with CIFAR dataset, and an accuracy of 99% with MNIST dataset.
- Comprehensive Data Analysis and Model Evaluation: Conducted extensive exploratory data analysis (EDA) and batch processing for efficient data handling. Meticulously plotted training, testing, and loss curves to evaluate model performance, enhancing understanding of neural network behavior and optimization strategies in image classification tasks.
Contact Me:
