Designed and deployed a PyTorch-based predictive model achieving 92% accuracy and reducing claim processing time by 30%. Cleaned and preprocessed 500,000+ claim records, extracting 25 features to improve model performance by 8%. Deployed the model on AWS and integrated it with claim systems for real-time predictions, cutting processing time by 25%. Conducted test/validation with 95% coverage, identifying improvements that reduced incorrect predictions by 5%. Set up automated monitoring and CI/CD pipelines (GIT, Jenkins) to reduce model drift by 12%. Built a real-time data pipeline using Apache Spark and Kafka handling 10,000+ events per minute, reducing latency by 40%.

Junior Data Scientist, Machine Learning at Space Infolab

January 1, 2021 - July 1, 2022

Developed a TensorFlow-based conversational AI chatbot achieving 90% user satisfaction and 25% reduction in support queries. Implemented NLP-based sentiment analysis and intent recognition, boosting responsiveness and accuracy by 15% and engagement by 10%. Preprocessed 17,000+ support interactions to guide chatbot development, increasing user satisfaction by 20%. Integrated the chatbot with 3 existing support systems, reducing session duration by 30%. Analyzed performance metrics to drive data-driven improvements (12% accuracy gain) and designed a knowledge graph with 4,500+ entities and 9,000+ relationships to enhance responses.