Pavan P.

Leading migration and optimization of machine learning models for customer propensity scoring and behavioral prediction.

Advisor: Prof. Sreyasee Das Bhattacharjee • Conducted tutorial sessions and provided one-on-one guidance to over 50 students, improving their understanding of SQL, ER modeling, and normalization. • Designed and graded hands-on projects involving database schema design, complex queries, and data integration. • Assisted students in setting up and utilizing PostgreSQL with PGAdmin, Tableau, and Power BI for practical assignments and data analysis. • Collaborated with faculty and fellow GTAs to develop course materials, ensuring a balance between theoretical concepts & real-world applications.

• Boosted revenue by onboarding 50+ global merchants and preventing $2M in fraud losses through ML-based fraud detection models. Designed decision tree and anomaly detection techniques to analyze transaction data and identify fraudulent patterns • Generated and Automated reports on Tableau Dashboards to conduct trend analysis and research on real-time payment volumes by capturing fraud beforehand leveraging merchant retention by 70%. • Implemented Google’s TabNet with LIME & SHAP for fraud model explainability. Boosted fraud rejection precision from 55% to 67% by identifying key features in 250K+ transactions. • Designed an end-to-end pipeline by extracting data from Cassandra DB, performing data cleaning, analysis, feature engineering and automating visualizations for ML models, deriving actionable insights • Led product innovation in two emerging markets—food and grooming—via market-specific ML strategies. Boosted revenue growth and improved market entry success through primary research and fraud insights.

Advisor: Dr. V Srinivasa Chakravarthy Thesis : Vision-based Navigation using DRL Techniques • Engineered a virtual environment in Gazebo and ROS software with a team of 4, enabling a bot to autonomously navigate towards a target object while circumventing obstacles. • Presented and Designed a Deep Reinforcement Learning Model by using DQNs & LSTMs, resulting in 22% reduction in navigation time with a 76% precision in expected path. • Engaged with over 20+ undergraduates to evaluate and assess their strengths in Python course, resulting in a 30% improvement in submission rates.

• Collaborated with a Senior ProductManager to deploy the RKGE recommendation algorithm leveraging Knowledge Graphs, achieving a 30% increase in recommendation accuracy and a 20% boost in user engagement metrics • Generated efficient path-based algorithm for KPRN recommendation model and evaluated with enriched MovieLens 1M dataset for TopK recommendations. Achieved Prec@10 (Top 10) with a 0.086 score. • Deployed in Samsung chatbot for customer service serving 100k customers.

Part of the Indigenous 5G Test Bed project that closely resembles a 5G deployment. Implemented RRC Reconfiguration module on 5G Wireless Communications following the 3GPP documentation. Encoded and decoded the RRC messages using BER and PER encoding rules in C++ for measurement configuration and Measurement report triggering.

Graduate Teaching Assistant under Prof. Sreyasee Das Bhattacharjee

Research Assistant under Dr V Srinivasa Chakravarthy Teaching Assistant under Prof. Harishankar Ramachandaran (Applied Programming Lab Python)