At Metafusion, I built and deployed an NL-to-MongoDB query generation model for real-time video data analytics. Fine-tuning, DPO, quantization, few-shot prompting, and caching helped improve its accuracy to 97% while keeping inference under a second on 8-core, 16GB systems.
At R2G Logistics, I developed and integrated A* and Christofides route optimization algorithms, reducing average delivery times by 25%. I also streamlined pathfinding systems, reducing workforce training efforts by 70% through simpler navigational logic.
At IIT Jodhpur, I developed a web-based course planner for faculty to manage schedules, lectures, and assessments. It includes timetable creation, syllabus tracking, and automated progress updates.
In an AI Society project, I developed a handwritten digit classification model using CNNs and the MNIST dataset. I’ve also designed a file compression and decompression tool with customizable compression levels and support for multiple file formats.

