Chintan Sumaria, Ph.D. Email & Phone Number

Washington, DC, US

Newark, NJ, US

Boston, MA, US

• Team: Catalog Data Science
• Implemented a multi-task evaluation workflow for Foundation Model (FM) representations (or embeddings) using a linear probe approach.
• Automated and scaled workflow to millions of product-images and hundreds of nodes using Google Cloud Vertex pipelines, enabling iterative FM development, achieving a reduction in execution time (to test a new FM) of.
• Conducted experimentation on a Sort Experimentation Framework (SEF) by exposing substitute SKUs on Store Front to customers to assess and compare substitution models.

Boston, MA, US

• Enhanced deep learning models using big data tools as PySpark and GCP Dataproc to predict SKU performance at launch ("Day Zero"), achieving an AUC improvement over 40%.

Plantation, Florida, US

• Team: Supply Chain Data Science, Supply Chain Research & Planning
• Part of a two-member team that created and developed an optimization-based Order Routing System 2.0 (ORS 2.0) that dictated shipping decisions for over 10 million packages and saved ~36M$ annually compared to its.
• Led a team of ~6 Data Scientists to create monitoring solutions, implement new service features, and to troubleshoot issues in production.
• ORS 2.0 was highly effective in saving costs and improving customer experience and was called out by Chewy's CEO in his letter to shareholders as a "proprietary machine-learning-driven software, which streamlines order.

Plantation, Florida, US

• Created a perfect replica of the existing ORS (ORS 1.0) to simulate the entire network.
• Created multiple GUI applications for internal customers using Flask.
• Performed various simulations to improve the core and pharmacy outbound Supply Chain networks.
• Optimized Cross-Dock network by formulating a Linear Programming routing strategy for vendors, and then by simulating the network for all vendors via stochastic Monte Carlo simulations, followed by a genetic search.
• Forecasted hour-level outbound volume for the entirety of Chewy network.
• Authored or co-authored various White Papers documenting history, limitations, improvement options, potential savings, future state, modes of operation, models, etc for existing and/or future versions of ORS.

Redmond, Washington, US

• Non-FTE role (via Soothsayer Analytics LLC)
• Developed MVP Data Science solutions for Microsoft C+AI Security and Azure Supply Chain using both Supervised (Classification, Regression, Forecasting) and Unsupervised (Clustering, PCA, NLP) ML algorithms.
• Applied machine learning to classify KPIs, detect anomalies, forecast costs, and analyze confidential HR data.
• Deployed models using Docker and utilized the Kusto big data platform.

New York, NY, US

• Image Entropy QuantificationApplied random distortions to the ImageNet dataset to fine-tune VGG16 using GPU-intensive instances on AWS in order to quantify entropy in images.
• Clustering Pharmaceutical Drugs by Side EffectsScraped Drugs.com, among other sources, to compile a list of drugs and side effects; Applied NLP techniques along with Word2Vec to cluster drugs into groups found to be.
• Analysis of LendingClub DataAnalyzed millions of records from peer-to-peer lending firm LendingClub to predict defaults; Created a Flask-based prediction tool.
• Regression Modeling for Craigslist Auto DataScraped and analyzed tens of thousands of Craigslist posts to understand the used car market.

Chicago, IL, US

• 400+ citations
• Carried out experiments in multidisciplinary areas of chemistry to calculate, predict, design, and test catalyst characteristics such as acidity, enantioselectivity, etc.
• Used regression models to predict and measure acidities of unknown compounds by computational and experimental methods.
• Performed quantum mechanical calculations or simulations of equilibrium molecular geometries (with/without solvation models) to study correlations between existing catalysts.
• Deciphered 1D/2D spectroscopic data generated by Nuclear Magnetic Resonance spectrometers.
• Generated visualizations of siRNA-protein complexes from the Cambridge Structural Database (CSD) using Python (pymol) (predoctoral work, review article published).

Doctor Of Philosophy - Phd, Chemistry

Coursework (In-Person) In Statistics & Microeconomics

Dual Degree (5-Year Integrated Msc)