At Zestgeek Technologies, I develop production-grade voice AI agents with LiveKit and Pipecat, achieving under 200ms latency. I’ve built call orchestration workflows and multi-provider LLM integrations with fallback chains.
I designed infrastructure for more than 30 concurrent voice conversations per pod using Docker, Kubernetes, and cloud platforms. Connection pooling and caching reduced operational costs by 40%.
At vTech Solution Inc., I designed an AI-driven data labeling and access authorization pipeline using prompt-based classification, Snorkel weak supervision, and Hugging Face models. I also built governance automation for metadata extraction, policy tagging, and retention enforcement.
I built SuperDialog, a conversational-AI runtime for real-time voice agents, and an open-source playbook skill for Claude Code and Codex. My work also includes evaluation tools for voice agents and a published engineering deep-dive on production playbook design.
