Daksh Jain Email and Phone Number

Bid & Budget Optimizationo A/B tested & developed phased-based optimization framework to allocate product bids & budget for search & browse ads on the Tokopedia platform, the framework utilized the existing constraint optimization model along with heuristics for product initialization, product selection, product rotation and order optimization which resulted in a +10.78% increase in impressions per shop, a 2.56% increase in order per shop & +11.03% increase in budget absorptiono Productionized & implemented heuristic-based rules by incorporating buyer relevance signals to produce optimal bids at product-keyword and product level for search ads which resulted in a +2.31% increase in ads revenue per impression, a +4.97% increase in ads budget absorption, 2.78% increase in ad orders per impression o Developed a regression model to predict CTR at product level for external ads platform which had 25% lesser mean absolute error than baseline using several slots, product, shop, and category attributes where forecasted CTR was used to convert CPC to optimal CPM bids to compete in open RTB auction. [POC]o Implemented A/B test framework for shop sampling & post-experiment analysis Developed a machine learning framework and identified key actionable insights to increase seller adoption for organic sellers to Tokopedia ads system by using exploratory data analysis & uplift modeling/causal inference techniques which helped acquire ~200 sellers per month. [POC]

Identified customers at risk of churning from e-commerce channels with key controllable/uncontrollable drivers and their quantitative and directional impact by developing a classification model using random forest & logistic regression having ~4 million transaction data and ~750Kcustomers [POC]

Productionized & developed a machine learning pipeline on Azure ml workspace to forecast net revenue, net revenue margin & shipments at monthly level having mape ~5% by utilizing several univariate and multivariate forecasting techniques, for multivariate forecasting scraped several macro-economic variables from Fred, Yahoo Finance & ADP websites. Migrated SAS pipeline to Python and deployed the same on Azure ml workspace, through Azure migration and code parallelization pipeline run time reduced by 97%. Also developed a Power BI dashboard to guide regional managers to track the efficiency of their service centers.Forecasted certain financial accounts at the day level for the current month and next three months by utilizing some statistical models using Linear Regression & ARIMA having mape <10%. These forecasts were used to calculate expected operating income by the finance team by using forecasts of revenue and expenses.Developed an OCR model to identify PDF bills using TensorFlow object detection and then extract useful information using Microsoft computer vision API. This initiative resulted in securing new projects from the same client. [POC]