Ashwin Ramanathan

Revenue Generation
• $10M USD in incremental revenue generated for non-fuel retail business, by virtue of data analytics enabled category management (Snacks, Beverages etc.) in Shell fuel stations, across the globe.
• $8M USD generated separately using promotions as a key lever for generating sales, includes development of a statistics-based tool for analyzing promotions across key categories.
Business problems solved
• Analytics based category management, using power BI and databricks. Specific areas include basket affinity, pricing, promotions, product placement and performance measurement.
• 5P analytics as a theme (pricing, products, promotions, placement, performance) for 360-degree category management.
• Customer churn from shell EV sites, based on customer lifetime data.


Artificial intelligence
• Developing an AI agent, capable of crunching monthly retail numbers and sending insights to stakeholders, in the form of a concise newsletter.
• Natural Language understanding chatbot, capable of answering business questions from the user. Based on a predefined data model.
• Completed POC of using Databricks Genie, AI playground, as viable alternatives to carry about agent creation, other AI initiatives.

Technology Development
• Enterprise data platform migration from HANA system to Azure data lake. Completed all necessary QC checks and migrated almost 100TB of data.
• Pioneered use of databricks, ML-Flow and other CI/CD ways of working in the wider team.
• Worked with IT in adaptation of unity catalog, as a governance mechanism for data layer.

Mentorship and guidance
• Mentored 50 + data scientists and junior members on the adaption of GitHub, VS code as ways of working.
• Conducted classroom training on using ML-Flow for running experiments, feature engineering, model selection and hypotheses testing.

• Accelerated revenue by developing and owning Market Basket Analysis solution for large scale retail data basket affinity analysis, on Azure Databricks platform.
• Conserved dashboard reading time by building Automatic Insights generator as for executive summary from long/detailed Power BI dashboards using R in backend of Power BI for data summarization.
• Succeeded in designing Out of stock identification tool- early warning system for retail store managers to better stock their inventory using Time series modeling in Azure Databricks.
• Engineered recommendation system, responsible for boosting indirect channel sales (B2B) of lubricants ,engine oils and coolants by virtue of giving apt recommendations for sales pitches based on customer preferences and buying history.
• Headed project for finding effect of customer satisfaction on sales volumes, managing end to end process from collecting data to developing regression models while getting feedback from stakeholders.
• Redesigned large scale claims processing filed by clients of shell, by developing automation frameworks focusing Alteryx workflows, thereby reconciling many productive man-hours.
• Tracked data retrieval time using parallel processing/multicore processing techniques in R, Python and implemented better ways to mine data from HANA database setting aside traditional SQL based querying

• Boosted customer trust on Amazon.com, by identifying and blocking fake and abusive product reviews, that spoiled overall customer experience, using traditional logistic regression modeling.
• Quantified success of product launches by conducting statistical significance tests on random sample of test and control group of customers.
• Designed Python program for monitoring and maintaining advertiser universe on Amazon.com, ensuring seamless journey tracking of prospective clients from lead stage to onboarding. This tool was used by sales executives and senior leaders alike.
• Ensured flawless data migration from traditional Oracle database to cloud based AWS redshift, by taking care of data quality, partitioning and phased onboarding of users onto new DB, thereby ensuring business as usual for all internal stakeholders.
• Engineered databases that addressed reporting and analysis needs of product managers and business analysts by clear understanding of business requirement, identification of source data in various pockets, building pipelines for data transfer and taking responsibility for any breakage or downtime in newly constructed DB, by drafting SOP.
• Contributed to operational excellence in data engineering by developing guidelines for efficient query writing, constructing ETL jobs and setting up of alarms to proactively fix leakages in data pipeline.

• Assisted tenured professors in operations and management divisions of university with their research work, in capacity of data analyst for fetching data , coming up with interesting patterns and gathering data evidence for given hypothesis that would help advance their research towards paper publication(s) in reputed journals such as INFORMS,IEEE.
• Researched public transportation data in New York city and examined distribution of commute between systems subway, buses, bike share and ride hailing apps such as uber and how this has possibly changed demographics of NYC and its carbon emissions.
• Studied time difference in open source software development by individual repositories versus organizational repositories on GitHub by understanding data behind commits, pull requests and open issue management

• Implemented automatic document classification of financial research papers, based on pre defined taxonomy using standard techniques in Natural Language processing such as LDA,LSA and N-gram modeling .
• Explored topic modeling toolbox API, built at stanford university as possible solution to achieve higher accuracy in multi -class document classification.

• Created sentiment analysis engine in Python for extracting and analyzing tweets on commercials that were aired in between NFL Superbowl, improving social media campaigns by 50 percent in terms of user engagement
• Delivered insights on user churn in websites, using traditional techniques in regression such as CART, thereby capturing shift in user behavior using clickthrough rate of different parts of website.