RAGURAM K A

Designed and implemented a virtual teacher using a 3D human mesh with real-time voice synthesis, leveraging NVIDIAs Audio2Face, Tacotron2, vision models, and pose detection algorithms to monitor student attentiveness, while incorporating diffusion models for enhanced interactivity and functionality. I have reduced Heygens re-talking models inferencing time by 30 percent. Developing an AI-based DDoS attack detection model for servers, utilizing Autoencoders and Digital Fingerprint analysis models from NVIDIA Morpheus to enhance cybersecurity.

Responsible for developing Python-based prompt and response system for training Large Language Model. Reducing model convergence time by 5 percent and enhancing overall performance. Developed and optimized Python scripts for data analysis, visualization through graphs, and automating various tasks to enhance efficiency and insights.

Designed and implemented a Python algorithm to correct GPS misreadings, leading to a 45% reduction in data inaccuracy by fine-tuning the underlying ML model responsible for data calculations.

Led the development of a video KYC system in compliance with RBI regulations, implementing an OCR model with 95% accuracy for text extraction and a facial recognition model with 98% accuracy from videos, streamlining client onboarding by verifying the authenticity of user-provided documents through video verification. Enhanced Food Pandas model for food item identification in images.