Projects

Developed a data-driven SQL model to enhance the SBA loan approval process. The project utilized historical data to analyze key factors such as loan amount, business location, loan term, and issuing bank, providing actionable insights for lenders. Key business questions included identifying factors that influence loan repayment, understanding how business location and loan terms impact approval rates, and assessing the likelihood of defaults based on different loan amounts. The primary goal was to predict loan repayment likelihood, enabling informed decision-making that supports responsible lending practices while increasing access to capital for qualified small businesses.

• View project

Developed a real-time audio analysis system using an Arduino Nano BLE Sense and implemented a K-Nearest Neighbors (KNN) machine learning model to detect coughing sounds. The project involved capturing audio data via the onboard microphone, preprocessing it to extract relevant features, and training the KNN model to classify audio patterns associated with coughing. The model was optimized for deployment on the resource-constrained microcontroller, ensuring efficient processing and real-time response. This system demonstrates the potential for low-cost, portable solutions in health monitoring, particularly for applications like tracking respiratory conditions or detecting symptoms in public health settings.

• View project

In this project, the focus was on applying Python for data science tasks, where key skills were developed to handle real-world data using Python's rich ecosystem of libraries. The project involved working with datasets, performing data cleaning, exploration, and analysis, and applying statistical methods to draw insights. Tools such as Pandas, NumPy, and Matplotlib were used for data manipulation and visualization, while Scikit-learn was leveraged to implement machine learning algorithms for prediction and classification tasks. Throughout the project, emphasis was placed on understanding the data science workflow, from data preprocessing to model evaluation. The goal was to develop proficiency in using Python for analyzing data, making predictions, and deriving actionable insights, while gaining practical experience that is directly applicable to real-world data science challenges.

• View project

I developed the Wallet Wise PowerBI Dashboard to monitor telecom revenue assurance and fraud management across multiple telecom services, including money, data, SMS, and voice accounts. The dashboard integrates real-time data from telecom systems to track key performance indicators (KPIs) and detect potential fraud, ensuring revenue integrity and minimizing financial losses. I designed the system to provide detailed insights on account usage, identify anomalies in revenue flows, and flag unusual patterns that could indicate fraudulent activity. It helps in monitoring customer behavior, visualizing service usage trends, and allowing for proactive measures to mitigate risks. By consolidating data from various telecom services, the dashboard enhances decision-making, enabling quicker responses to emerging issues and optimizing revenue collection processes.

• View project

In this project, a ResNet-18 deep learning model was used to classify a dataset of approximately 3,000 Chest X-Ray scans into three categories: Normal, Viral Pneumonia, and COVID-19 for early COVID-19 detection. The dataset was preprocessed by resizing, normalizing, and augmenting the images to improve the model's performance and generalization. The ResNet-18 architecture, known for its residual connections, was fine-tuned for this specific classification task. The model was trained using supervised learning and evaluated with metrics like accuracy, precision, recall, and F1-score to ensure robust performance, especially in distinguishing COVID-19 from viral pneumonia. Data imbalance was addressed using techniques such as data augmentation and transfer learning. The final model was deployed as a diagnostic tool, aiding healthcare professionals in early detection of COVID-19, enhancing the efficiency of screening processes in clinical settings.

• View project