- Objective: Enhance user experience by recommending personalized products tailored to their specific sport, location, and other pertinent factors1) Designed a framework that enables seamless integration of new recommendation algorithms, reducing development time from weeks to 2-3 days while significantly enhancing code readability 2) Implemented multiple recommendation and personalized floors for both the app and web platforms, tailored to customers at various stages of their journey, yielding an approximate 7% increase in CTR3) Leveraged advanced algorithms, including Matrix Factorization and RNN to build personalized floors based on customer’s purchase history4) Built a CI/CD workflow to automate the deployment process. Integrated recommendation responses with a serverless API using AWS Lambda, enabling seamless access for diverse teams and channels

- Objective: The project aims to identify the specific sports each user practices by ranking their interactions within the sports environment1) Built a data ingestion pipeline to ingest data extracted from different source systems into Databricks Delta Lake2) Leveraged Pyspark to create a workflow implementing multi criteria based decision making methodology (TOPSIS) to rank user's sport preferences based on their purchasing and event booking behavior3) Automated the entire pipeline execution using Pyspark, ECR, AWS MWAA (Airflow), incorporating data quality checks on input, intermediate, and output tables, reducing runtime from 4-5 hours to 45 minutes

- Objective: To obtain the missing RI benefit and maintain data integrity, the company attempts to recognise and remediate the miscoded claims.1) Developed Date of Loss & Loss Cause models to classify claims as natural hazard/working loss by using rule based and machine learning approach that saved ~$83 millions2) Re-trained the existing Date of Loss and Loss Cause models to capture recent natural events3) To streamline the pipeline, Date of Loss model was reformed to be based on Python from SQL that saved ~50% of manual effort4) Assisted the team in performing and preparing the remediation files in excel for the claims assessed which is later consumed by Claim Center team

- Objective: Develop an AI enabled market intelligence framework by integrating internal and external data sources to predict market price and demand movement of high impact commodities1) Created an AI enabled market intelligence framework by integrating internal and external data sources to predict market price and demand forecast in over 40+ countries2) Created and automated HQL scripts for data cleaning and pre-processing3) Developed a forecasting framework to anticipate the price and demand of petrochemical products for different countries using an ensemble of time series and regression4) The product developed stimulated the time taken in decision making by 90% thus enabling client to sell products at the highest price point

- Objective: Develop AI chat-bot for one of the largest petrochemical retailers to help them uplift organisational efficiency and reduce the resolution time from hours to real-time.1) Drove business conversations during the requirement gathering, project road map development, stakeholder management phases to ensure quality end product2) Developed a web crawler module in python to automate the process of scraping product and document level data from website in a clean, structured and efficient way 3) Designed product ontology that enables fast and effective query processing on Neo4j by writing python scripts that transform scraped data to W3 standard knowledge graph4) Wrote exhaustive unit tests for web crawler, ontology creation modules and performed stress, load tests on the product using Apache Jmeter5) Assisted the development of chat-bot using RASA open-source (NLP DIET classifier, flash text) which helped in addressing the customer's queries from hours to real time

- Objective: Develop an attribution framework and generate campaign analytics report for all the Digital Media Campaigns.* Digital Marketing Attribution (Analysis) | Home Improvement Retailer : 1) Developed a solution to drive vendor funded personalised digital campaigns for re-targetable customers by designing a data driven digital monetisation strategy2) Analysed re-targeting campaigns using Multi touch attribution on top of Omniture data leading up to 40% increase in channel optimization3) Created an attribution logic to rightly credit campaign performance across all channels for selected brands to showcase the capability of retailer's reporting in yearly brand meet4) Created a framework to assess the efficiency of statistical models such as Markov Chain & Shapely on multi touch customer journey data5) Offered personalised support to various brands by providing real time insights and recommendations and attained a net profit of ~$100 million* Digital Marketing Attribution (Data Engineering) | Home Improvement Retailer : 1) Developed Apache Airflow data pipelines to automate the queries for generating the vendor reports that in turn reduced the process time by 50% and saved ~$5500 per month2) Implemented validation framework in Airflow to run the checks on input, intermediate and output tables that reduced the reporting variation by over 40%