Built a sophisticated RAG chatbot specializing in tax applications for various products. The chatbot's natural language processing capabilities allow users to query complex tax scenarios, receiving instant, context-aware responses that factor in jurisdictional nuances and product-specific tax rules.Engineered a dynamic KNN/LLM based router that intelligently directs queries to the most suitable language models based on prompt complexity. Aligned language models with clinical datasets to achieve state-of-the-art (SOTA) results across various models, including 3B, 7B, 8B, and 70B. This alignment has paved the way for impactful advancements in the healthcare domain.Designed and trained an advanced deep learning model for brain tumor detection and classification using MRI data. Leveraging Phi-2 LLM, this project not only aids in early diagnosis but also provides personalized treatment instructions for better patient care.Merged multiple 7B models using innovative techniques such as Dare-ties and Slerp to enhance model performance. This effort resulted in a remarkable average score of 76.66 on the Open LLM leaderboard, showcasing substantial improvements in model capabilities.Developed a scalable cloud-based microservices solution for photo gallery management on Google Cloud Platform (GCP). This solution incorporated auto-scaling, monitoring, and logging features to ensure high availability and reliability. Rigorous load testing using K6 Cloud was conducted to validate system performance under varying conditions.Implemented a product categorization system from the ground up using Convolutional Neural Networks (CNN). This text-based classifier effectively categorizes products with an accuracy of 88%.