Project 1: IndiGo: (Industry | Aviation)Use case 1: R Scripts triggering | Cloud Composer1.Deployed 200 DAGs to copy R Scripts from GCS to Compute engine & trigger R Scripts in Windows server.2.Implemented a DAG generator DAG that binds Jinja template & config file to create multiple DAGs automatically using yaml.3.DAGs are configured to trigger emails based on success & failure.Use case 2 & 3: Database Migration & Archival (Compute engine (MSSQL) to BQ)1.Migrated 700+ tables from MSSQL to BQ as one time load.2.Deployed Composer Airflow scripts to archive 700+ tables every 15 days.3.Schema for every table is replicated from MSSQL to BQ.4.Composer DAG is scheduled to initiate a Dataflow job to fetch MSSQL tables & archive into BQ.5.Similarly Migrated 40+ tables from MySQL to BQ as one time load. (Cloud SQL (MySQL) to BQ)Project 2: Birlasoft (Haleon) (Industry | Consumer Health care)Use case 1: (SAP-SFMC to BQ)1.A composer DAG is deployed to run a batch pipeline everyday to pull data using SFTPtoGCS operator & write into BQ landing zone for multiple tables.2.BQ SP's are tasked in DAGs to move data in different layers of BQ. Landing to cleansed & curated zones in BQ.3.DAG is deployed to dynamically include tables. It also, handles data load failure & automatically load the failed date's data referring to a date in meta data table.Use case 2: (Abandoned carts | Near-realtime data transfer)1.Idle cart messages for more than 15mins are sent to PubSub by an API.2.This is received by a PubSub topic which triggers a Cloud function which formats & writes into BQ. Another function with same trigger to send message to SFMC for email.3.Data from BQ is sent to SFMC & DV360 through Cloud functions & ComposerBig Query - PII masking - (Personally identifiable information)Automatic scripts are created to apply PII policy tags for taxonomies.

Job Profile:Create GCP Architecture diagram - Create Machine Learning solution procedure - Create ML Data workflows - Create Data Engineering Pipelines - Create Inference pipelines - Test & Demo clients - Prepare TDD documents. (Leading Machine Learning, Platform Engineering, Data Engineering & Software Development teams)Project 1: Oscar - Health: (ICD10 code Extraction & Classification from Medical Claims)Technologies applied: 1.Extracted around 320K pdf files using Cloud Vision API & OCR processors.2.Data pipelines created with cloud composer & Apache airflow. 3.Used caml algorithm to train & develop PQR-AI (Payer quality & risk) model. Model is built using Cloud AI - running around 50 experiments with different data sets & batch sizes. Tuning hyper parameters, step size & sequence lengths using Tensor flow.Project 2: Walden University: (Text to Speech - Podcast Creation)Technologies applied:1.Web-scraped the university site periodically using selenium web-drives operating in GCP App engine.2.Text is summarized & coded to receive SSML files (speech synthesis markup language).3.SSML served as input to Azure-speech engine to generate podcast on different metrics like voice, pronunciation, tone & speed rate.4.Generated podcast is surfaced in podcast app for Walden students in production.Project 3: Hartford Insurance group: (Email & Document - Classification & Entity Extraction)Technologies applied:1.Extracted email & its attached documents text - using OCR processor.2.Classified them using XG-Boost model - classified emails are applied to entity extraction.3.ACORD - insurance domain specific (pre-released-processor) were used to extract entities from attached docs.4.Alternate logic was also used to script a “region of interest” over the docs to pick entities.

Job Profile:Create ML models - Write python scripts - create AWS data flows - Data science - ETL - Web scraping.Project 1: Corning: (Anomaly Detection - Classification)Technologies applied:1.Pipe lined data from a client database to preprocess & train data.2.Classified each transaction among 10 different anomalies as cost & compliance anomaly categories.3.Used multiple models while training - Light BGM, XG Boost & Random Forest. Validated model using Confusion matrix, Cap curve & regularization methods.4.Entire process occurs on EC2 machine which picks up pickled pre-trained model. Model results were surfaced on an application with its classification.Project 2: Corning: (Trending News - Suggestion)Technologies applied:1.Web-scraped news websites from a list of websites.2.Used tf-idf vectorization & naive bayes model to pick model relative words across websites for suggestion.

Job Profile:Analyze banking data - Write Python scripts - Create ML models - Create end-to-end data engineering pipelines - Design SQL tables.Requirement: Classify performing & non-performing assets (loan payers).Technologies applied:1.Performed descriptive & prescriptive analytics on features such as family, net worth, repayment delay, region.2.Pipe lined data from a database to preprocess & train data.3.Classified an account as performing & non-performing asset based on previous loan repayments using XG boost algorithm.4.Model was launched to production after testing one year old customer data.

Job Profile:Analyze Logistics data - Write Python scripts - Create ML models - Create end-to-end data engineering pipelines - Design SQL tables.Requirement: Cab allocation demand predictionTechnologies applied:1.Applied predictive analysis on cab allocation & routing data from corporate car booking domain data using pandas.2.Multiple features were created based on distance from office to home, region, traffic status during the time, car size & availability. 3.Random forest regression was used to find the value of optimal demand for a car to be allocated.

Job Profile:Data collation - Data Analysis - Prescriptive, Descriptive & Predictive, Implement data web apps - Create regression ML models - Write Python Scripts.Requirement: Data analysis & ERP implementation for IFM & real estate data.Technologies applied:1.Different types of data were analyzed using pandas, regex, charting methods. 2.Implemented multiple web applications & modules required for facilities management & real estate domains. Trained clients & site teams on usage & tracking.