Working as part of the LEaD Program on various ML and AI projects

Grading and assisting the students for Machine Learning class CSCI- 635

o Launched animal detection notification methodology for Advanced Driving Assistance System using the YOLOv5 instance segmentation model o ~10% boost in Mean Average Precision (MAP) of thermal imaging segmentation model using Genetic Evolution and Augmentation

Grading and assisting Masters student for Computational problem solving using python

Achieved 80% precision in identifying cardiac abnormalities via visual representation with the use of Manifold ML & Laplacian Eigen Mapso Converted a complex 918-point EKG signal to a single 3D visual heart simulation for medical imaging using MATLAB

o Saved $2MM(~10% reduction) marketing costs related to new title releases using data wrangling, feature engineering and statistical modelling approach on demographics, transactions, & browsing attributes using SQL, Statsmodels and Pythono Engineered a 20% uptick in click-through rates via production ready Lookalike model powered by DBSCAN & XGBoost on Data bricks & AWSenhancing programmatic marketing campaign prioritization for Disneyo Managed a team of 3 to streamline marketing insights pipelines using PySpark, SQL, Tableau resulting in a 30% reduction in the turnaround time

o Improved performance 3x and accuracy (F1 score ~ 67%) of credit history assessments for loan applications by extracting image details using YOLOv3,Optical Character Recognition (OCR) & Named Entity Recognition (NER) techniques from govt. authorized identification cardo Integrated a Logistic Regression model for a financial collection agency to indicate debt delinquency risk probability. Employed propensity modeling, Population Stability Index, & characteristic analysiso Managed a team of 2 to develop extensive Extract, Transform, Load (ETL) pipeline and visualization using SQL, Pandas, Python, SAS and Tableau

Churn Prediction for a major US cable television provider/multiple system operator using Deep learningon Google Cloud PlatformBusiness Problem:Client was facing a issue of customers churn. They were interested in incorporating deep learning for obtaining better KPIs.Used python andSQLto query data out of Bigquery(GCP)and do data wrangling and imputations of missing features.Extracted data out of GCS into AI Notebooksto analyze customer behavior, demographics, products and performedEDA.Built and trained a Wide and Deep neural network using tensorflow to predict churners. This information was conveyed to the business for win-back.

Targeted Marketing and Messaging Solution for a major US cable television provider/multiple system operatorBusiness Problem:Client wanted a B2B solutionusing data science for effective targeted marketing and messaging solution.Performed EDA on demographics, viewership,customer behaviordata. Did feature selection to identify features for model training.Built and Trained a Random Forest for lookalike modelingto obtain the target household out of population dataset.Wrote code to perform Principal Component analysis for clustering channels together based on the viewership of the targeted household to obtain the primary watched segment of each householdNBO(Next Best Offer) solution for major Telecom provider based in Belgium Business Problem:Client had the key objective of transforming the (Offer) simulation solution for its B2B customers for mobile product domain,encompassing new, Winback and Retention.Used R for segmentation of MSISDN numbers based on customer and product levelsApplied time series forecasting to forecast the Data, Voice and SMS usage for next 30,60,90 days based on the past data of each MSISDN number using algorithms like ARIMA. This usage was used to forecast the billsUsed collaborative filteringbased on clustering techniques like PAM and distance measures like Gower's distance to cluster MSISDNs by taking the user's product and their segment information to find three best offers for each MSISDNs.

Worked as an trainee engineer on SDH technology and PDH networks .Experienced in installing and commissioning Huewai MUX 500,550,X8,X16. Have idea of traffic shifting and swaping procedures