- Developed a machine learning model using PyTorch to teach a small robot to navigate a room and avoid obstacles by predicting potential collisions.
- Collected and labeled training data to train a neural network, ensuring accurate predictions for obstacle avoidance.
- Designed, trained, and tested a neural network for real-time collision prediction, improving the robot's decision-making and movement efficiency.
- Implemented and evaluated the model in a simulated environment, optimizing network performance for generalization in new scenarios.
- Designed and implemented the applicaZon using LangChain, showcasing experZse in chains, agents, document loaders, text spli_ers, output parsers, and memory management.
- Integrated third-party APIs (ProxyURL, SerpAPI, Twi_er API) to scrape data from the internet, enabling the applicaZon to find LinkedIn and Twi_er profiles based on user input.
- Developed and deployed AI-powered applicaZons leveraging Azure AI Studio and advanced Large Language Models (LLMs), incorporaZng custom Python funcZons and APIs to automate workflows.
- Used Retrieval-Augmented Genera[on (RAG) to incorporate diverse data sources including PDF, Excel, and txt.
- Combined LLMs with Azure’s vision, speech, and document intelligence services to enhance user interacZon.
- Developed an ASP-based program using Clingo to optimize insurance case assignments to referees, ensuring workload balance, region and case type matching, and cost optimization.
- Implemented hard constraints to ensure workload limits, region, and case type matching, and damage thresholds were met.
- Incorporated weak constraints for cost optimization, payment fairness, workload balance, and referee preference respect.
- Identified and resolved issues related to incomplete knowledge base information during testing phase.
- Demonstrated ability to handle complex decision-making processes accurately through ASP-based solution.
- Developed a recognition system integrating supervised and unsupervised machine learning techniques for analyzing time series data from two asynchronously operated Medtronic 670G systems collected at 5-minute intervals over a 7-day period.
- Implemented adaptive data cleaning and extracted features, including Fast Fourier Transform (FFT) and Entropy calculations, as well as time span, from the time series dataset.
- Distinguished meal and no meal time series data through the training and testing of machine models, employing sklearn k-fold cross-validation for robust model training.
- Trained Support Vector Machine (SVM) and Decision Tree Machine (DT), achieving a noteworthy DT F1-score of 77% and an Accuracy of 81%. Extracted ground truth and performed clustering using DBSCAN and Kmeans, attaining minimal DBSCAN entropy of 0.22 and maximum DBSCAN purity of 0.83.
The primary goal of this project was to provide data-driven insights for the UVW marketing team to help them better understand their target market and support them in making informed decisions. Various exploratory data analysis methods were used to analyze the relationship between variables and locate the critical factors affecting individual salaries.
- Developed complex SQL queries using JOINs, CTE (Common Table Expressions), and subqueries to analyze job market data, identifying top-paying data analyst roles and the most in-demand skills in the industry.
- Leveraged advanced SQL functions like GROUP BY, COUNT(), AVG(), and ORDER BY to aggregate and rank data, revealing high-demand skills associated with higher salaries for data analysts.
- Optimized SQL scripts for performance and readability by utilizing window functions, temporary tables, and effective indexing strategies, enabling efficient data extraction and insights generation from large datasets.
- Led a team in building React Web App, focused on optimizing customer satisfaction.
- Developed robust backend APIs using Node.js/Express and the Sequelize, ensured seamless database operations and efficient CRUD functionality through rigorous testing.
- Introduced user-friendly filters for customized search experiences and improved overall usability through intuitive UI enhancements.
- Enhanced visual appeal and streamlined image retrieval through AWS S3 integration.
Web.App.Demo.mp4
- Built a custom eCommerce platform with React, Redux, Node, Express and MongoDB, featuring product search, carousel, pagination and more.
- Authenticated with JWT & HTTP cookies, admining customers, products and orders.
- Full featured shopping cart with PayPal & credit/debit payments.
Deployed on Render: https://mars-visit-application.onrender.com
This project is a web application built using the MERN stack (MongoDB, Express, React, Node.js). The application is designed to handle multi-step form submissions with validation and data persistence. The data is collected from users and stored in a MongoDB database via the backend API. The form includes personal information, travel preferences, and health and safety details.
- Multi-step Form: A step-based form process to collect user information.
- Form Validation: Validates form data for required fields, email and phone formats, date consistency, and other custom rules.
- Data Persistence: Stores user submissions in a MongoDB database.
- Responsive Design: Built with React-Bootstrap for a mobile-friendly UI.
- API Integration: Connects the frontend to the backend API for data submission and retrieval.
- Developed ASP.NET SOAP web service and client for API calls and data processing.
- Developed an iOS app in Swift with MVC architecture, featuring an intuitive user interface for data input, information retrieval, and web page access.
- Employed Azure Virtual Machine for VS, facilitating cross-platform development.
Project.Demo.mp4
- Conducted object-oriented analysis and design (OOAD) to develop a sports concussion assessment application, focusing on requirements elicitation, system object identification, and user scenario analysis to enhance athlete health monitoring.
- Designed and documented UML artifacts including use-case diagrams, CRC (Class-Responsibility-Collaborator) diagrams, class diagrams, sequence diagrams, and state diagrams to model system structure and behavior, ensuring comprehensive understanding and effective stakeholder communication.
- Implemented a Java-based console application that enables athletes to input symptoms, generates symptom summaries, and provides sports medical practitioners with tools for risk assessment, including severity score calculations and condition tracking.
- Developed a user-friendly interface for athletes to enter symptom evaluations and for practitioners to review symptoms, calculate severity differences between games, and receive risk indicators for better decision-making and athlete management.
- Leveraged tools like Astah and Draw.io to create detailed object-oriented models and visual representations of system components, facilitating efficient software development and system architecture planning.
- Applied Java programming and OOAD principles to create modular, reusable software components, enhancing system maintainability and scalability, while adhering to software engineering best practices.
- Refactored a legacy Java system by addressing object-oriented design violations and eliminating content, common, and control coupling issues, significantly enhancing code quality, maintainability, and modifiability.
- Applied UML class diagram analysis to identify design flaws and restructured key components to promote low coupling and high cohesion, improving system scalability and flexibility.
- Implemented the Factory Pattern in the PersonnelFactory class to manage different types of personnel, providing a scalable architecture that supports future extensions and reduces dependencies.
- Encapsulated data and logic by making class fields private and introducing getter and setter methods, ensuring proper information hiding and compliance with encapsulation principles.
- Developed and extended system functionalities, including personnel management features (adding, searching, and displaying personnel), ensuring robust performance and alignment with software development life cycle (SDLC) practices.
- Designed a robust and extendable system to determine whether vehicles in Lyft’s new rental fleet should be serviced.
- Drafted a UML class diagram representing a new reorganized architecture.
- Refactored the codebase with Factory and Strategy design pattern for multiple product creation and manipulation.
- Implemented unit tests and added new functionality using test-driven development.
- Developed measurable and testable quality attribute scenarios for a client-server web application, focusing on critical factors like time behavior, confidentiality, and recoverability to ensure system reliability and security.
- Created comprehensive software architecture documentation using the C4 modeling approach, including system, container, and deployment diagrams, to visualize components and interactions, supporting effective software architecture implementation.
- Conducted performance testing, security testing, and recoverability testing to validate system resilience under various conditions, ensuring adherence to industry standards for confidentiality and minimizing downtime to under 10 minutes.
- Utilized tools like Draw.io to construct C4 model diagrams, enabling clear communication of architectural decisions and ensuring alignment with software development life cycle (SDLC) best practices.
- Prioritized software quality attributes in the architectural design process to enhance system maintainability, scalability, and usability, contributing to a robust and reliable online shopping platform.reation and manipulation.
