Utkarsh Maheshwari
Data Scientist
I’m a graduate student at Georgia Tech with 4 years of industry experience in analytics and applied machine learning. Before Georgia Tech, I worked at Genpact, ZS Associates, and Samsung Research focussing on commercial analytics for life sciences and technology industries. My work included demand forecasting, market sizing, go-to-market strategy support, and operational planning, where data and models were used to guide real business decisions. During my master’s, I’m building hands-on projects in machine learning, deep learning, and time series forecasting with a focus on practical applications.
Projects & Research Work
NutriLens
Applied MLEnd-to-end ML system to predict food processing levels based on its ingredients and recommend healthier alternatives with similar taste profiles. The core classifier achieves 97% accuracy, with fine tuned LLM-based explanations added to help users better understand nutrition labels. Deployed as a web application using a Flask backend and React frontend.
WhyHere
Applied MLSmart neighborhood recommendation system for Atlanta that ranks areas based on user priorities like safety, transit, rent, and amenities. Uses collaborative filtering to suggest neighborhoods liked by similar users and a distance-decay scoring model enhanced with custom feature weighting to provide explainable rankings. Integrated data from 12+ APIs and processed it using Databricks and PySpark. Visualized results with an interactive map.
Lucid-GAN
Research PublicationNovel GAN architecture for enhanced image inpainting. Restores damaged images and upscales quality. Published at CIVEMSA Conference 2021.
Speech Enhancement System
Research PublicationConvolutional encoder-decoder architecture for speech enhancement. Published in LNEE 2021. Developed at Samsung Research Institute.
Deep Convolutional GANs
Paper ImplementationImplemented the original GAN paper by Ian Goodfellow to generate new images by learning the distribution of a dataset. Trained convolutional generator and discriminator networks to create realistic images from random noise.
Neural Style Transfer
Paper ImplementationImplemented the original neural style transfer paper to apply artistic styles from one image onto photographs while preserving content structure. Used CNNs to extract content and style representations and generate stylized images.
Professional Experience
Assistant Manager at Genpact
Jan 2025 – May 2025Led sales force optimization projects, building predictive models, SQL pipelines, and Power BI dashboards to drive data-informed decision-making for US pharmaceutical clients.
Associate Consultant at ZS Associates
Jan 2024 – Jan 2025Managed end-to-end analytical workstreams including demand forecasting, financial modeling, and launch analytics for major pharmaceutical clients using Python, SQL, and Power BI.
Associate at ZS Associates
Jun 2021 – Dec 2023Built demand planning and financial models to support demand forecasting and launch strategy decisions for major pharmaceutical firms using Excel, Python, and VBA enabling data-driven decision making.
Research Intern at Samsung Research Institute
Oct 2020 – Mar 2021Developed deep neural networks using TensorFlow for speech signal denoising, performing data wrangling and feature extraction.