Sales Insights Dashboard
Interactive Power BI dashboard offering a comprehensive view of key sales metrics, enabling the identification of trends, patterns, and business opportunities. Built in association with the American Red Cross using DAX, data modeling, and end-to-end data cleaning and visualization.
Bank Marketing Campaign — EDA & Prediction
Developed a predictive model for ABC Bank to optimize marketing for a new term deposit product. Analyzed customer data from past interactions to forecast subscription likelihood, helping the bank focus resources on high-probability customers. Evaluated models with and without the 'duration' feature to ensure practical real-world applicability.
AI-Driven Clinical Decision Support for Diabetes
Conceptual research project designing and evaluating an AI-driven clinical decision support system (CDSS) for diabetes management. Assesses effectiveness across clinical outcomes, patient-centered measures, health-system impact, and ethical considerations using prospective, real-world, and equity-focused evaluation methods.
Diabetes Readmission Prediction
Led comprehensive Exploratory Data Analysis to uncover patterns and address data quality issues — outliers, missing values — in a hospital readmission dataset. Identified critical factors influencing readmission rates for diabetes patients and provided actionable recommendations to improve healthcare outcomes and ML predictive accuracy.
Fake News Detection with Diverse Model Approaches
Explored fine-tuned Pretrained Language Models (PLMs), Large Language Models (LLMs), and traditional ML models to classify real vs. fake news. Evaluated model performance using three datasets including Liar and WELFake, comparing effectiveness across approaches for misinformation detection.
TriMet Data Pipeline
Built a data engineering pipeline to extract, clean, validate, integrate, and visualize TriMet bus data for the Portland metro area. Configured a GCP Linux VM to parse breadcrumb data into Kafka topics, built a Confluent Kafka consumer for daily ingestion into a database, and created a secondary pipeline integrating transit stops with breadcrumb data.
Binary Classification for Diabetes Detection
Performed data cleaning and handled missing values to prepare a diabetes dataset for ML model training. Employed multiple machine learning techniques and evaluated a multilayer perceptron against a Bayesian classifier, comparing their performance in predicting diabetes based on diagnostic measures.
Health Database from U.S. Census Data
Designed and built a health-domain database using United States Census Bureau data. Enabled users to analyze how different demographic groups were affected by the COVID-19 pandemic, providing data-driven insights into the pandemic's impact across various populations.