I'm Premith, I Specialize in developing cutting-edge AI technologies and innovative applications, I bring a blend of research acumen and practical engineering to solve complex problems. I have current and previous previous work-experiences at ABB, Virginia Tech, Samsung R&D Institute, Supervue Ai, and Zebi Data. My passion lies in building AI products and services to enhance business processes, data privacy, process automation, and user experiences.
Let's connect and explore how we can leverage AI to create transformative solutions. Open to opportunities that challenge the status quo and drive innovation.
Work Experience
Primary inventor of a patented technology: Dynamic Auto-Calibration System, IP India Patent App No: 202041034063
Research Papers
&
Patents
Inventor of a Collaborative Patent: n-Neighborhood Temporal Feature Transformation between Samsung R&D and Andhra University, App No: 202241043300
Author of Generative Design For Manufacturing: Integrating Generation with Optimization Using a Guided Voxel Diffusion Model, Published in AAAI Conference
Author of AMMDAS: Multi-Modular DAS with Generative Masks Processing, Published in IEEE Access Journal
Author of l, r-Stitch Unit: Encoder-Decoder-CNN Based Image-Mosaicing, Published in IEEE Access Journal
Author of Attention Based Multi-Patched 3D-CNNs with Hybrid Fusion Architecture, Published in JCC Journal
Author of Generating Optimized 3d Designs for Manufacturing Using a Guided Voxel Diffusion Model, Published in MSEC Conference
Recent Projects
LLM Powered Multi-User Video Conferencing Application with Automated Meeting Minutes
Jan 2023 - May 2023
Developed Python, JS, and WebRTC based video conferencing app leveraging GPT API for automated meeting minutes generation (ASR-transcribed meeting audio). Resulted in 30% time savings. Used Express JS for secured HTTP requests to stream app media.
Zero-Shot Video Question Answering Using CLIP, CLAP, and Mistral Models [Demo]
Sept 2023 - Jan 2024
Developed a Video QA app by integrating Mistral-7B LLM-based RAG pipeline on audio & video context extracted using CLIP, Grounded-SAM, & CLAP models. Outperformed 10x larger Flamingo-80B model on multiple Video-QA benchmarks
