Rakesh Chowdary Vunnam

Built a multi-camera exercise analysis system using YOLOv8 pose estimation and SVM classification to automatically detect exercise repetitions and compute 200+ biomechanical metrics across front, side, and diagonal views for objective performance assessment. Developed V-DWT-JEPA2 architecture with LLM decoder for advanced sports video understanding, and created a domain-specific chatbot with post-training QA optimization using LLaMA family models. Designed a multi-agent architecture enabling real-time coordination between video analytics and coaching components; deployed via Azure cloud-native services with containerized microservices, and implemented an LLM-based evaluation framework with guardrails for safety and correctness.

Led cross-domain AI initiatives spanning healthcare, manufacturing, and marketing. Built agentic RAG systems with multi-modal embeddings and reranking, implemented video-to-document pipelines and browser extensions for real-time knowledge capture, and deployed scalable ML pipelines on AWS/Azure with governance and auditing. Architected autonomous agents for real-time coaching insights and integrated LLM-driven safety and compliance checks.

Developed end-to-end data analytics solutions, created advanced DAX calculations, automated data refresh pipelines, and built real-time IT service dashboards. Implemented row-level security and governance, led migration of legacy reports to Power BI, and produced mobile-optimized dashboards for on-the-go decision making.