Air Quality Prediction Model 

Air pollution poses a serious environmental and public health challenge, particularly in rapidly industrializing regions of China. Accurate forecasting of air quality is essential for effective pollution management and policy intervention. However, existing prediction models often fail to capture the complex spatial and temporal dependencies within multi-source environmental data.

To address this limitation, this study proposes an Attention-Enhanced Spatio-Temporal Graph Convolutional Network (AE-STGCN) that integrates spatial–temporal attention mechanisms with graph convolutional learning. The model fuses heterogeneous datasets—including air pollutant measurements, meteorological variables, and urban point-of-interest (POI) data—into a unified grid-based feature representation. By dynamically learning the significance of different spatial regions and temporal intervals, AE-STGCN enhances predictive accuracy and interpretability.

Video based obstacle avoidance using Deep Reinforcement Learning (DRL)

This ongoing research focuses on developing an intelligent obstacle avoidance system for autonomous mobile robots using real-time video input. The project integrates computer vision and deep reinforcement learning (DRL) to enable robots to perceive and navigate complex dynamic environments efficiently.

A YOLO-based object detection model is used to identify and classify moving obstacles, while DRL algorithms such as DQN and PPO are applied to optimize navigation strategies through continuous learning and interaction with the environment. The entire system is developed and tested in the ROS and Gazebo simulation platforms, allowing for experimentation under variable conditions like sensor noise, lighting, and motion dynamics.

The goal of this research is to create a robust end-to-end framework that fuses visual perception and decision-making, enabling autonomous systems to adaptively respond to real-world scenarios with improved safety, precision, and reliability.

API Testing Framework

This project is a Python-based automated testing framework built using Pytest to validate RESTful APIs. It provides a complete testing pipeline — from executing API test cases to generating structured JSON reports and interactive visualizations of the results.

The goal is to make API quality assurance efficient, repeatable, and data-driven. The framework not only verifies API responses but also gives teams a quick, visual understanding of test outcomes and performance trends.

Client Automation Test

The Client Automation Test Framework is a Python-based automated testing system designed to validate client-side applications and APIs through scripted test cases, automated execution, and dynamic reporting.

It ensures functional reliability, performance stability, and data accuracy of client-facing systems by automating repetitive testing tasks, minimizing manual intervention, and producing detailed analytics for each test cycle.

This framework provides an end-to-end automated testing pipeline — covering API validation, client workflow verification, and visual reporting.

E-commerce Checkout API Performance Testing

The E-Commerce Checkout API Performance Testing project focuses on evaluating and optimizing the performance, scalability, and reliability of an e-commerce checkout service.

This project simulates real-world user behavior — such as adding products to the cart, calculating totals, applying discounts, and completing payments — and measures how efficiently the backend API handles varying loads.

The objective is to identify bottlenecks, measure response times, and ensure that the checkout system can handle high traffic without performance degradation.

QA Practice Documentation - E - commerce Checkout API

This project focuses on testing and documenting an E-commerce Checkout API using both manual and automated QA methods. It covers the full QA lifecycle — from requirement analysis and test planning to automation, performance testing, and reporting.

The goal was to ensure the checkout workflow (cart, order, payment) runs smoothly, efficiently, and securely. Using tools like Pytest, Postman, Docker, and Plotly, the project validates API reliability and visualizes test outcomes for better decision-making.

Classifying Fashion MNIST images WITH CNNs

Convolutional Neural Network (CNN) to classify clothing items in images using the Fashion MNIST dataset. The process includes data loading, preprocessing, model creation, training, evaluation, and visualization. The dataset was normalized and split into training, validation, and test sets to prepare it for CNN training. A custom CNN architecture was designed and trained using the Adam optimizer and sparse categorical cross-entropy loss function. The model achieved an accuracy of approximately 90.66% on the test dataset.