top of page

Introduction

me.jpeg
Atharv Yeolekar

Data Scientist / Machine Learning Engineer

Email:

Location:

Atlanta, GA

Organization:

Walmart Global Tech

A Bit About Me

Hi there! I'm Atharv Yeolekar, a Generative AI Data Scientist with a passion for developing cutting-edge machine learning models and big data solutions. I've had the pleasure of working at Walmart Global Tech, Grubhub, and as a researcher in a NASA and NSF-funded lab.

Through this blog, I aim to document my journey of learning and mastering Large Language Model concepts. I'll be sharing insights, discoveries, and practical applications in the form of a blog/journal. Join me as I explore the fascinating world of AI and break down complex topics into digestible and engaging content.

Let's embark on this learning adventure together!

Work Experience

September 2023 - Current

August 2022 - August 2023

August 2021 -  July 2022

  • Developed time series-based anomaly detection for brand performance metrics, identifying significant deviations.

  • Led the development of a graph-based Product Association submodule, projected to generate $55M incremental profit.

  • Scaled the Product Association submodule using PySpark and GCP, processing over 200 million items to enhance dynamic product bundling.

  • Optimized supplier strategies through feedback from major suppliers like P&G, Unilever, and Nestle.

  • Developed GrubLM (Grubhub’s Large Language Model) using BERT and RoBERTa, harnessing food data to generate refined embeddings for various downstream tasks.

  • Created a pioneering algorithm leveraging Greedy KMeans and Boosting techniques to streamline CRM campaign targeting, enhancing profitability.

  • Developed, deployed, and optimized a lifetime orders (LTO) prediction model on AWS using Jenkins and Azkaban, resulting in annual savings of $1.4M.

  • Collaborated on a NASA and NSF-funded project, focusing on multivariate time series-based feature selection algorithms for space weather data.

  • Implemented graph-based and class-separability feature selection methods to improve solar flare classification performance, contributing to advancements in space weather prediction models.

bottom of page