Hrutvi Barad

Data Scientist with 2+ years of experience in retail and insurance domains, working on customer analytics, risk modeling, and predictive machine learning solutions.

Experience in analyzing large-scale structured datasets to derive insights on customer behavior, transactions, and claims patterns. Strong background in building and deploying machine learning models for classification, regression, clustering, and forecasting use cases.

Experience in applying AI/LLM-based solutions using OpenAI APIs and Hugging Face transformer models for sentiment analysis, text classification, and unstructured data processing. Hands-on experience in customer segmentation using clustering techniques such as K-Means for marketing and risk targeting. Developed predictive models for churn prediction, claim probability, demand forecasting, and risk scoring applications.

Proficient in performing end-to-end data preprocessing including missing value treatment, encoding, scaling, and outlier handling. Skilled in exploratory data analysis (EDA) and statistical methods to identify trends, correlations, and business drivers.

Experience in evaluating model performance using metrics such as ROC-AUC, precision, recall, F1-score, RMSE, and cross-validation. Built dashboards and visual reports using Power BI, Matplotlib, and Seaborn to communicate insights to business stakeholders. Collaborative experience working with marketing, product, and underwriting teams to translate business problems into data-driven solutions.

Data Scientist at The Home Depot (2025-08 – Present)

• Analyzed customer transactions and behavioral data using Python, Pandas, and NumPy to identify purchasing patterns, product affinity, and lifecycle trends.
• Designed customer segmentation models using K-Means Clustering to improve targeted marketing and campaign effectiveness.
• Developed predictive models for churn prediction, demand forecasting, and purchase propensity scoring using Scikit-learn.
• Applied machine learning algorithms including Logistic Regression, Random Forest, KNN, and Gradient Boosting for classification and prediction tasks.
• Performed pricing sensitivity analysis using price elasticity modeling to evaluate demand response across product categories.
• Integrated OpenAI API-based text analysis workflows to perform sentiment summarization on customer reviews, enabling faster identification of product issues and improving feedback analysis efficiency for business teams.
• Experimented with OpenAI embeddings for semantic similarity matching, grouping similar customer feedback and product complaints to enhance categorization and insight discovery.
• Built reusable data pipelines and automation scripts for reporting and model execution.
• Conducted EDA and statistical analysis on customer demographics to uncover behavioral drivers.
• Developed dashboards using Power BI, Matplotlib, and Seaborn for business insights.
• Collaborated with marketing and product teams to translate business requirements into analytical solutions.

Data Scientist at IAT Insurance Group (2024-06 – 2025-07)

• Worked with insurance claims and policy datasets using Python to analyze risk patterns, claim behavior, and loss drivers.
• Conducted EDA (Exploratory Data Analysis) to evaluate claim frequency, severity, and underwriting risk factors.
• Built predictive models to estimate claim probability and support risk scoring using Scikit-learn.
• Developed classification models using Logistic Regression, Decision Trees, Random Forest, and XGBoost (basic tuning).
• Improved model performance using feature engineering on policyholder, customer, and claims data.
• Handled preprocessing including missing value imputation, encoding, and feature scaling.
• Applied PCA (Principal Component Analysis) for dimensionality reduction in high-dimensional datasets.
• Optimized models using GridSearchCV for hyperparameter tuning.
• Integrated a Hugging Face transformer-based LLM (BERT/RoBERTa) to perform NLP-based analysis on insurance claim narratives, improving automated categorization of claim types and supporting faster risk classification.
• Applied LLM-based text embeddings using Sentence Transformers (Hugging Face ecosystem) to cluster similar insurance claims and identify hidden patterns in high-risk policyholder segments.
• Identified high-risk customer segments using historical claims and behavioral trends.
• Created visual reports using Matplotlib and Seaborn to support underwriting decisions.
• Documented insights to assist business stakeholders in risk-based decision-making.

– University of North Carolina at Charlotte (2024-05)