Erica Wilson Joseph

Highly motivated engineer with diverse experience in the tech industry which includes data analytics,software quality.

Successful at leveraging data trends and patterns in producing intelligence to enhance business planning and requirements

1. Analysis on Seller segregation using Machine learning model:

• Used SQL to query data from multiple databases and performed ETL and data processing techniques on the real time company data.
• Performed different approaches to use input features for unsupervised Kmeans clustering model to analysis the data effectively.
• Used data visualization techniques to provide insights to the leadership derived from the analysis. The analysis provided insights to increase opportunities in field of high value sellers versus low value sellers and coming up with solutions for retention.

2. User Engagement Analysis for an e-commerce enterprise using SQL (Academic)

• Performed analysis for the count of user engagement across months, weeks, and days using SQL and Python scripts in the Databricks platform.
• Delivered visual representations to highlight the user trends and possible reasons for the drop in the user activity and dug deeper into the data using SQL queries to provide insight into the most impactful features.
• This analysis suggested the areas that would need improvement to engage and retain the users.
• Link : https://databricks-prod-cloudfront.cloud.databricks.com/public/4027ec902e239c93eaaa8714f173bcfc/7818844666170531/643399434729642/6814813109746479/latest.html

2. Generating GUI Test Scenarios for Intelligent Mobile Apps using ML models, NetworkX Python package (Academic)

• Collected and built a custom dataset which included Mockup diagrams with annotations of GUI elements and GUI elements, which were to be trained based on their AI functionality.
• Used YOLO-V5, EfficientDet, and Detectron2 models for the classification of GUI elements on each screen of the AI application and further detected the relationship between two screens and a single screen.
• Determined the link classification using python code to determine if the relationship between the two given screens was AI or Non-AI screen.
• In the integration stage, used the Network X python library and Depth First search algorithm to generate the possible test scenarios based on the outcome of the GUI detection model results, relationship model results, and link classification results.