• Developed Classification model using Logistic Regression to predict the Out of Stock of an Item based on Perpetual Inventory.• Developed Prophet model to forecast net sales at vendor and category level for vendor negotiations and estimations of future trends.• Implemented MLFlow capabilities for machine learning models in order to productionize from end to end.• Worked on data pipeline to pull source records for every one-minute refresh using kafka streaming.• Worked on optimization of ML model jobs to reduce overall execution time and optimal utilization of clusters in databricks.• Integrated Machine Learning models in Python API applications for real-time inference.• Transformed and deployed machine learning models into scalable and maintainable production environment using CI/CD setup Jenkins (Decommissioned) and Github Actions.

Responsibilities• Implemented various classification techniques to predict the fraud or fake customer applicant using LSH (Locality Sensitive Hashing) methodology based on similarity probability and achieved model accuracy of 89%, which saved the business 1.8 million dollars.• Implemented DBSCAN cluster technique to identify different groups of customers based on their purchasing behavior and achieved 84% accuracy.• Built model pipelines to orchestrate data flow using PySpark, Kafka, Hive, MongoDB in Azure Databricks. • Trained classification model using different algorithms Logistic Regression, Random Forest Classifier, SVM, Multinomial Naïve Bayes, XGBoost Classification and deep learning models using TensorFlow. • Developed the collaborative filtering model to build the recommendation system using surprise python library.• Developed complex SQL queries by using advanced SQL techniques including complex joins, subqueries, views, indexes, stored procedures, etc. to implement business logic.• Used graphical packages which produces ROC AUC Curve to visually represent True Positive Rate versus False Positive Rate. Performed Lift and Gain analysis of classification models to evaluate model performance.• Implemented CI/CD data Pipelines using Azure Devops to deploy algorithms into production and monitor models

Responsibilities: • Used predictive analytics such as machine learning and data mining techniques to forecast company sales of new products with a 92% accuracy rate.• Built and trained a deep learning model using TensorFlow on the retail upgrade data, and reduced wafer scrap by 12%, by predicting the likelihood of wafer damage. A combination of the z-plot features, image features (pigmentation) and probe features are being used.• Built Tableau reports and dashboards to find areas of improvement for sales resulting in $420, 000 in annual incremental revenue. • Used 100K+ sentiments to understand user sentiments over the time. Data was facilitated from various sources such as consumer receipts, questionnaire, retailors, surveys, etc.• Improvised streamlining processes, resulted in reduction time of processing by 40%.• Applied Statistical Methods, classification algorithms to study customer satisfaction and brand reputation while improving compliance with recommended maintenance. • Developed an POC to implemented Autoencoders to identify anomaly detection with an accuracy of 83% and model was approved to be used in production. • Collaborated with data engineers and operation team to implement ETL process, wrote and optimized SQL queries to perform data extraction to fit the analytical requirements.• Performed data analysis by using Hive to retrieve the data from Hadoop cluster, SQL to retrieve data from MySQL.• Implemented Partitioning, Dynamic Partitions and Buckets in Hive for efficient data access• Created and implemented various shell scripts to automate the jobs.

Responsibilities• Enhanced the SCM module analysis which helped the client to improve the user satisfaction rate for annual maintenance contract by 7%.• Automated the reconciliation of data from different sources which helped bring down the financial closure time for client from 10 days to 3 days. • Automated the processing and integration of millions of records of transactional data from Oracle EBS and legacy systems to improve real-time reporting of product metrics and reduced the downtime. • Handled imbalance problem in datasets using under sampling and over sampling SMOTE techniques.• Identified best tuning techniques using Grid Search CV, Cross validation techniques to Outage problem by building best ensemble algorithms Random Forest, Gradient Boosting, XG Boost and evaluation metrics WMAPE, MAPE, MSE, MAE.• Experienced in Artificial Neural Networks and Deep Learning models using Theano, TensorFlow and Keras packages.• Used stored procedures, triggers and views to provide structured data into actionable items in order to accommodate the user requirement.• Challenge was to validate data in Analytical platform frequently. Implemented Automation pipeline while handling 16Tb of data on daily basis to perform QA/QC.• Extracted data from various web services APIs to compare data using web scraping python, cleaned data and built dashboards using PowerBI to communicate analysis to business leaders.• imported and exported data between HDFS and Relational Database Management systems using Sqoop.• Brought data from various sources into Hadoop and Cassandra using Kafka. • Implemented CI/CD Pipelines using AWS Devops to deploy data pipelines into production and monitored them.

Problem was to classify Auto insurance holding customers to send promotions, provide offer on exclusive roadside assistance services based on customer’s car brand, number of cars owned, enrollment of roadside assistance program in last 10 years, credit history, education level, job profile and income level. Solved With the help of large datasets machine learning models and trained to further identify potential customers to provide roadside assistance service offers.Technologies: Python, Statistics, Machine Learning, Tableau, SQL