Feb 2022 – Feb 2024 · 2 yrs · Bangalore Urban, Karnataka, India

• 𝐏𝐫𝐨𝐣𝐞𝐜𝐭 – 𝐌𝐚𝐫𝐤𝐞𝐭 𝐌𝐢𝐱 𝐌𝐨𝐝𝐞𝐥𝐥𝐢𝐧𝐠- 𝐌𝐌𝐌
• 𝐎𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞: Build a MMM model on the advertising spends data using regression for optimizing spends on media
• channels and analyze the impact of macroeconomic and other factors on the sales.
• Performed the data acquisition from GCP and client database, data preprocessing, data validation and
• analysis.
• Applied geometric adstock transformation on the media variables to introduce the advertising properties
• (e.g., decay, lag-lead) into the model. Used Facebook Prophet library to extract monthly
• trends.
• Built nested regression model to encounter issue of collinearity. Predicted store sales with 0.93 R2 score.
• Performed decomposition of the variables to understand the incremental effects of each media spends and
• external variables on sales, visualized using waterfall chart.
• Automated adstock hyperparameter tuning and deployed MMM nested model on local machine using
• streamlit.

Nov 2020 – Feb 2022 · 1 yr 3 mos · Pune, Maharashtra, India

• 𝐏𝐫𝐨𝐣𝐞𝐜𝐭: 𝐉𝐨𝐛 𝐭𝐢𝐭𝐥𝐞 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧 (𝐍𝐋𝐏 𝐂𝐥𝐚𝐬𝐬𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 𝐩𝐫𝐨𝐛𝐥𝐞𝐦)
• Given the features like Job description, domain info, job level, country, and other relevant features, performed NLP and feature engineering to use the text data information in predicting job title.
• Build and compared Logistic regression and tree-based ensemble models, selected XGBoost model which gave strong probability values for class labels.
• Built end-to-end ML pipeline using MLflow Recipe, for model tracking and registering the validated model.
• 𝐏𝐫𝐨𝐣𝐞𝐜𝐭 – 𝗖𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝘀𝗲𝗴𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻 𝘂𝘀𝗶𝗻𝗴 𝗰𝗹𝘂𝘀𝘁𝗲𝗿𝗶𝗻𝗴 𝗯𝘆 𝗶𝗻𝗰𝗼𝗿𝗽𝗼𝗿𝗮𝘁𝗶𝗻𝗴 𝗖𝗟𝗧𝗩 𝗮𝗻𝗱 𝗥𝗙𝗠 𝗮𝗻𝗮𝗹𝘆𝘀𝗶𝘀.
• 𝐎𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞: Categorize the customers into High, Medium and Low value to drive the marketing efforts to maximize retention rate and acquire new customers.
• Performed data preprocessing and feature engineering to extract features like recency, frequency, monetary value (RFM), average order value, CLTV score, etc.
• Applied K-means clustering algorithm on the data with all the necessary features to segment the customers.
• Delivered results using dashboard and performed monitoring and retraining of the model.

Jan 2020 – Nov 2020 · 10 mos · Navi Mumbai

• 𝐏𝐫𝐨𝐣𝐞𝐜𝐭: 𝐅𝐫𝐚𝐮𝐝 𝐜𝐥𝐚𝐢𝐦𝐬 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐢𝐧𝐭𝐞𝐠𝐫𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐭𝐡𝐞 𝐟𝐥𝐚𝐠𝐠𝐢𝐧𝐠 𝐬𝐲𝐬𝐭𝐞𝐦 𝐢𝐧 𝐭𝐡𝐞 𝐬𝐨𝐟𝐭𝐰𝐚𝐫𝐞.
• 𝐎𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞: Classification of the fraudulent claims and flagging them in the initial stages of claim process.
• Developed understanding of business problem and contributed in designing project stages.
• Performed Data acquisition, preprocessing, EDA, and feature engineering to derive new features from
• existing. Used transfer learning using BERT to extract word vectors for text data i.e., cause of accident.
• Extracted the data using SQL and preprocessed the big data using parallel computing systems like Dask.
• Trained Logistic regression as base model and finalized Random Forest after comparing other models.
• Generalized the model by testing it against the test queries to make the model production ready.
• Contributed in end-to-end deployment in AWS Sagemaker, created endpoint and carried out API testing in Postman.
• 𝐏𝐫𝐨𝐣𝐞𝐜𝐭: 𝐒𝐞𝐠𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐥𝐨𝐰, 𝐦𝐞𝐝𝐢𝐮𝐦 𝐚𝐧𝐝 𝐡𝐢𝐠𝐡 𝐥𝐢𝐚𝐛𝐢𝐥𝐢𝐭𝐲 𝐜𝐥𝐚𝐢𝐦𝐬 𝐮𝐬𝐢𝐧𝐠 𝐜𝐥𝐮𝐬𝐭𝐞𝐫𝐢𝐧𝐠.
• 𝐎𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞: Segmenting the claims in low, medium and high liability claims to maintain equal workloads amongst surveyors as well as assigning high liability claims to external surveyors.
• Developed understanding of business problem and performed data acquisition, data preprocessing and EDA.
• Used domain knowledge and feature engineering techniques to derive the new features from existing.
• Used K-means clustering to cluster the claims in low, medium and high liability claims.
• Developed a rule-based system with the help of IT team to maintain the workload within surveyors.

Aug 2018 – Dec 2019 · 1 yr 4 mos · Mumbai, Maharashtra, India

• 𝐏𝐫𝐨𝐣𝐞𝐜𝐭: 𝐅𝐫𝐚𝐮𝐝 𝐢𝐧𝐬𝐮𝐫𝐚𝐧𝐜𝐞 𝐜𝐥𝐚𝐢𝐦 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐚𝐥𝐞𝐫𝐭𝐢𝐧𝐠.
• 𝐎𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞: Classification of the fraudulent claims and integrating it with app that will send notification
• alert.
• Performed data extraction from company’s database using MySql, stored and maintained updated data.
• Performed preprocessing like, removing duplicate and missing value rows etc., using Python and Excel.
• Completed basic EDA using Python and Tableau and presented the insights of the data before team.
• Performed other supporting tasks to build and productionize the model.