ML Models for Google Ads Ranking

You already know.www.mistatlanta.com#getmistified

Generating Weak Supervision Machine Learning models to enforce Google Ads policies for image-based ads.

- Built a scalable and secure Data Lake to centralize siloed EC2 internal data using AWS Lake Formation and a variety of other native AWS solutions.- Orchestrated ETL workflows to extract, transform and load real-time data from AWS data stores such as AWS DynamoDB, RDS (MySQL), S3 and Athena.- Deployed ETL Workflows using AWS Glue to execute Python Apache Spark Glue jobs and crawlers to extract cross-regional and cross-account data from 150+ internal AWS production service accounts.- Automated the creation of 25+ nested CloudFormation stacks that update and maintain region-dependent resources using Python and YAML.- Engineered a VPC (Virtual Private Cloud) Network in conjunction with an Elastic Network Interface to tunnel network traffic to and from internal secure AWS networks. - Presented AWS QuickSight Analytical dashboards to discover hidden trends and outliers and to drive better business decisions and insights.- Completed an internal Amazon Machine Learning Accelerator (MLA) program to certify best ML practices at Amazon with AWS SageMaker.

• Worked with Machine Learning engineers to create ML models to improveinventory record inaccuracy detection by 75%.• Utilized XGBoost and Scikit-Learn for designing Supervised Machine Learningmodels and Keras for Feedforward Neural Networks to predict inventorydiscrepancies.• Orchestrated ETL (Extract, Transform, Load) pipelines with Apache Airflow andqueried daily database with Google BigQuery.• Conducted exploratory data analysis with several Python libraries includingPandas, Numpy and Matplotlib.• Integrated Google Cloud Platform services for Data Analytics - Composer andBigQuery, Storage - Cloud Storage and Computation - Kubernetes Engine.

• Created database application with SQL & Python to perform measurementsand analyses across 360 million records on private healthcare database.• Collaborated with data science team to develop and test performancealgorithms in Python• Created visualizations of database-query results to assist backend team oninternal data processing.• Structured, cleaned, and performed analyses of healthcare database, workingin Linux environment.

• Worked alongside Production Managers and Sales Specialists to createvisualizations of low voltage products.• Managed SAP database of low voltage products and directly assistedcustomers in product selection.• Wrote bash scripts to assist Open Orders Team to improve project and filemanagement.