- Utilized OpenAI's GPT-4 API to tailor responses for specific contexts, significantly enhancing user interaction with about 90% accuracy in data interpretation.- Developed a groundbreaking application that interprets PDF content, leveraging adjusted frequency and presence penalties in OpenAI's API to deliver near 100% relevant and precise answers. This application was used to develop a solution in the ESG Hackathon.- Engineered a sophisticated vector database for Retriever-Augmented Generation (RAG), ensuring efficient storage and retrieval of custom documents, thus dramatically boosting data accessibility and processing speed.- Created an innovative application that translates prompts into SQL queries using GPT-4, revolutionizing business analytics by providing deep insights directly from the database.- Successfully implemented LoRA (Low-Rank Adaptation) on the Llama 7B model, achieving results on par with RAG, thereby demonstrating exceptional skill in model optimization and AI efficiency.

- Engineered an innovative AR Visualizer App to effectively break down and demonstrate the core steps of Marker Based AR, harnessing the capabilities of Python, Tkinter, and OpenCV.- Designed an interactive Pathfinding Visualizer along with a dynamic Snake Game, both driven by prominent pathfinding algorithms like BFS, DFS, and A*, implemented through Python and pygame.- Took the lead in conducting classes on topics like marker-based AR and planar AR. Employed custom-built visualizers in Python to make intricate technical terms more accessible and engaging.- Established myself as the go-to liaison between a group of 15 students and faculty, consistently demonstrating prowess in swiftly resolving both software glitches and curriculum-related concerns.