Developed a machine learning model to predict and categorize RTO rates into risk levels, enabling the identification of high-risk customers using techniques such as XGBoost and CatBoost, leading to a 40% increase in seller retention.Developed a Hierarchical Product Category Prediction Model by experimenting with transformer-based models like BERT and a customized XGBoost algorithm, successfully implemented across multiple companies including Shiprocket, Pickrr, and Omuni.

Achieved a remarkable reduction in cost per claim from 70Rs to 15Rs through meticulous data analysis and optimisation techniques by employing diverse statistical modeling approaches like outlier identification and transactional risk scoring, leading to monthly savings exceeding 1Cr+.Pioneered the adoption of groundbreaking industry-first techniques, such as syndicate analysis, to unveil correlations among disparate individuals and spike analysis methodologies to detect abrupt and anomalous spikes in claims activity.Actively contributed to the development of a straight-through processing system for health insurance claims, resulting in a dramatic reduction in processing time from 72 hours to 5 minutes, significantly enhancing operational efficiency and reducing business costs.Integrated machine learning and forecasting algorithms into the claims adjudication system, enabling real-time fraud detection and decision making, resulting in a further reduction of fraudulent activities and ensuring the integrity of insurance claims processes.

Interned at IIT Patna for a Deep Learning project under the guidance of Dr. ChandranathAdak, focusing on a Deep Learning based study on Land Cover Classification.Conducted experiments involving standard models and performed a comparative analysis

Developed a low-cost PV Characteristic Plotter which helps to reduce efforts in analyzing defective panels by 80%.Innovative implementation of Multi-Camera Person Tracking and Re-Identification System, addressing the requirement for tracking across non-overlapping cameras for public safety. This solution eliminates manual tracking, resulting in a cost reduction of 90%.