@Microsoft: Detecting Tech Support Frauds

System detects and classifies Tech Support Frauds and sends data to the Smart Screen, Bing and Azure teams, and investigators

to block bad actors, bad resources and start law actions.

oDeveloped image and text classification pipelines (obtain data from scheduled ETLs, prepare it, feature engineering /

extraction, model training / verification, classification) (processed 100k images/day)

oDeveloped multilingual text classifier (for 9 languages) based on machine learning, NLP, boosting algorithms (with 97%

of F1 Score) (processed 100k text data/day)

oDesigned fault-tolerant system by saving current state and results

oDesigned fully autonomous system (get data, process it and classify, send results to the database and other teams)

oDecreased run time by implementing multithreading