Deepa Arul

Astute Data Scientist with data-driven and technology-focused approach. Communicates clearly with stakeholders and builds consensus around well-founded models. Excelled in transforming complex data into actionable insights, Successfully developed predictive models, Machine learning models and data-driven solutions to optimize business performance. Enhancing chatbot performance by engineering GenAI RAG models, Agentic AI and mentoring junior developers, showcasing exceptional analytical and leadership skills.

• Conducted correlation analysis on 2,500 test attributes, uncovering relationships and their impact on testbench performance.

• Developed an internal benchmarking tool using LangChain to evaluate and compare new LLMs, delivering detailed performance insights.

• Built a time series model to predict fault frequencies in testbench machines, enabling proactive maintenance and reducing downtime.

• Designed a time series model to forecast utilization trends for industrial washing machines, optimizing operational efficiency.

• Created a CNN-based image classification model to classify gear teeth as OK or Not OK, enhancing quality control.

• Developed a GenAI RAG-based chatbot using Azure services to automate document support, reducing manual workload.

• Implemented agents using AutoGen with sequential and nested chat actions to streamline complex workflows.

• Leveraged LLMs for automated data analysis, including descriptive insights, attribute evaluation, translations, and summaries.

• Built regression and logistic models to predict customer behavior metrics like sales, margin, trips, and quantity from Hand-Picked Offers (HPO) data, feeding into scorecards and charts.

• Developed 511 logistic models to predict customer shopping likelihoods within 7-day and 28-day windows, tailored to unique shopping patterns.

On-Shelf Availability (OSA)

• Created logistic models to predict item availability on store shelves, generating alerts to maintain inventory levels.

• Designed analytics processes to calculate inventory levels and update product availability on the Meijer product display webpage.

• Built forecasting models to predict product demand across shops, selecting the best models for training and scoring demand values.

• Distracted Driver Detection: Built a deep learning image classification model using artificial neural networks to identify distracted drivers.
• Sentiment Analysis on Amazon Reviews: Developed and tuned a machine learning model to predict sentiment scores from Amazon product reviews.
• Vintage Art Data Analysis: Conducted exploratory analysis on vintage art data using Tableau for insights and visualization.
• Dimension Reduction on Wine Dataset: Applied PCA for dimensionality reduction on the wine dataset as part of unsupervised learning.
• Customer Churn Detection: Identified potential churn for a mobile service provider using predictive modeling.
• Recommendation Systems: Designed product and movie recommendation systems to enhance user experience.
• Telecom Customer Churn Analysis: Analyzed customer churn in the telecom industry to identify factors driving attrition.
• Onion Sale Analysis: Visualized and predicted onion sales trends using data analytics.

PG-BDML