Sheikh Shahidur Rahman

Fraud Detection: I developed a supervised (XGBoost, GAN) and an unsupervised (ClusterCatch, Anomaly detection) solution in order to reduce abusive behavior of KDP (Kindle Direct Publishing) reading sessions in real time ($7.8M loyalty savings per year). Campaign Analysis: I developed a solution to compare eBooks in order to evaluate the performance of A/B testing. Used greedy search over K-medoid (variation of K-means) algorithm to produce similarity scores for different sets ($3M revenue lift per year).

• We created a methodology for headline suggestion for financial reports using LSTM (encoder-decoder) in NLP domain. I worked closely with analysts to automatically suggest headlines having higher catchiness index. •I worked with financial analysts to produce forecasting results for report publications. Created a Shiny tool in order to automate forecasting methods. •We developed an unsupervised clustering analysis in order to compare brand rankings across regions.

• Customer lifecycle management ( CLM ): We analyzed and built strategies (descriptive analytics) for prospect identification and customer retention (predictive analytics) in collaboration with Marketing team. • I collaborated with the marketing team & product team for their A/B testing . Automated the repetitive campaigns. • We analyzed and blocked Ola money (payment) fraudsters on our platform in collaboration with the product team & DevOps team.

• Walt Disney World in Orlando, Florida: We built a real-time recommendation (ALS algorithm) engine and realtime crowd management system (Shortlisted for Strata+Hadoop Conference, Presented by O'Reilly and Cloudera, London, May 5-7, 2015) • Cox Communications truck roll optimization: We built a predictive model in order to reduce truck rolls for system repairs.