Headline

Data Scientist | Serving Notice (Immediate Joiner) | AB InBev | Ex-Tiger Analytics

About

Immediate Joiner (Serving Notice) | Data Scientist with 7 years experience in solving complex business problems across varied industry verticals like Retail, Banking, Supply Chain, Telecom, Sales, etc, helping businesses achieve gains in terms of business performance, revenue, cost and time, and automatizing manual processes over 100s of GBs of data. Python | Statistics | Predictive Analytics | Machine Learning | Microsoft Azure | Big Data | SQL/PSQL/Oracle | MS Excel | SAS.

Location

Greater Bengaluru Area

AI stack

claude

gemini

ChatGPT

Work experience

Data Scientist at AB InBev, Nov 2022 - Present · 3 yrs 11 mos, Bangalore Urban, Karnataka, India, Product Purchase Propensity and Experiment Designing I E-commerce, Retail Analytics • Designed an end-to-end SKU recommendation solution for 20,000 stores countrywide, for internal app-based marketplace, achieving precision at 70% and recall at 90%, using a single XGBoost model. • Executed rolling window-based modelling and created features for last 90 days to predict the probability of purchase of each SKU on the 91st (present) day, with about 25 features used finally. • For each day, designed one single script to do historical features and target creation, daily training of the model, grid search-based parameter selection and recommendation generation. • Deployed the same in Databricks and Azure Data Factory, enabling daily automatic execution. • Designed 4 A/B testing experiments to test hypothesis like success of new SKUs, people-based marketing, etc. Portfolio Optimization Solution I Retail Analytics • Developed a SKU rationalization solution using mixed-integer programming (MIP) optimization to reject unprofitable SKUs basis their volume shares, production costs, sales intermittency, age, business strategy etc. • Designed a SKU cannibalization algorithm using Gowers distance to identify most similar SKUs basis pack type, price segment, target audience, flavor, VLC cost, etc for demand transfer of rejected SKUs. • Found customized similarity weights based on above and market share to estimate volume cannibalization and financial impact. • Ensured model & feature customizability and explainability for business by converting the solution to a product, enabling the testing of multiple scenarios using various feature sets.

Senior Analyst Data Science at Tiger Analytics, Mar 2021 - Oct 2022 · 1 yr 8 mos, Bangalore Urban, Karnataka, India, Cross-Sell Best Product Recommendation - Created an end-to-end solution to predict the next best product to existing customers for a multinational European bank across 11 categories. - Determined problem scope to update new product categories and redefined existing ones. - Built a propensity model to recommend best products to existing customers, achieving up-to 98% performance on business metrics. - Created multiple data extraction, feature engineering and scoring pipelines to deploy the solution in the production environment. Optimized Inventory Allocation - Created an inventory allocation application to automate a manual process for a leading manufacturing firm, reducing execution time from an hour per plant-product combination to seconds. - Created extensive data handling pipelines to determine demands, importance and inventory availability, and wrote python script to determine the optimized allocations. Business Query Classification - Implemented a coding framework to handle severely imbalanced data using multiple balancing techniques. - Ran multiple modelling algorithms to automatically select the best possible one, and used it for predictions.

Analytics Consultant at Emplay Inc., Apr 2020 - Feb 2021 · 11 mos, Bangalore Urban, Karnataka, India, - Deal Similarity Algorithm: An algorithm based on k-means clustering to shortlist successful deals closest to open deals, with the same logic extended to content and competitor recommendation. - Team Performance Algorithm: An algorithm to create team performance KPIs including team leader-member heirarchy definitions spread over combinations of leader role, year and three geography levels. (Run time ~ 60s) - Learning Chatbot: Coded quiz scoring logics based on SQL and information updation/editing commands in the chatbot, with normalized backend SQL tables.

Associate Analytics Consultant at Emplay Inc., Nov 2018 - Mar 2020 · 1 yr 5 mos, Bengaluru Area, India, Rep Coaching Algorithm & Chatbot Design: A large project spread over the following milestones: • Linear Regression & Aggregation based benchmarks prediction: Predicted personalized KPI targets for sales reps, enabling them to achieve their sales quotas, using linear regression and sampling averages. • Recommendation Engine: Built a recommendation engine to give personalized improvement strategies to sales reps, which handled varying performance scenarios, combining 7 data sources. • SQL-based Chatbot: Managed data & SQL based conversation flows while working in a team of 6 people. The bot involved SQL based data collection from user, information presentation to the user, large backend manipulations, dynamic dataset updating by the user in the backend, etc.

Data Science Intern at Emplay Inc., Aug 2018 - Oct 2018 · 3 mos, Bengaluru Area, India, Configuration-based KPI creation: Created a Configuration based code library for KPI calculation, to escape repeated calling of functions, usable across the organization.

Education

Data Analyst & Data Scientist, Dataquest.io

Bachelor of Technology, Mechanical Engineering, CGC College of Engineering, Landran