Barath Murugan

• Implemented SentenceTransformer model for efficient classification of Salesforce tickets for l4 and l5 issues.
• Fine-tuned DistilBERT for categorizing bug Jira tickets into major categories: bug, feature request, task.
• Developed tool for determining sample size for A/B testing using power analysis, including F-test, T-test, and non-parametric tests in Streamlit app.
• Created document-based LLM chatbot to generate context-aware responses for user inquiries.
• Conducted competitor analysis with linear regression model in Mexico region, evaluating ride volumes and payments against competitors.

Developed QueryBuddy tool to assist data analysis team with query summarization and optimization.

• Utilized generative AI to improve error identification and streamline table migration processes.

• Developed ResNet50 model for adult content detection Image classification (92% accuracy), replacing costly 3rd-party API, Cost savings 599$ per month.
• Built content Recommendation engine using TF-IDF + Cosine Similarity, and conducted A/B testing found that increasing avg session time from 2 to 11 mins.
• Created a Review Insight Analyzer for Shopify stores to derive actionable insights from customer reviews, boosting store optimization and customer experience.
• Developed an intelligent Rasa-based FAQ chatbot for Mybranz's mobile app, enhancing user engagement and support.
• performed AB tetsing using google firbase for AD varients to our clients and increaset CTR with new style CTA button statsitically proved .
• Built an Email Marketing Generator using GenAI, allowing stores to auto-generate promotional content for customer outreach and ads.

• Created predictive models for Fraudulent insurance claims using Random Forest, XGBoost, and Neural Networks (F1: 0.92, AUC: 0.96).
• Deployed real-time detection pipeline via Flask API; reduced false claim payouts by 22%.