Vaibhav Chaudhary

I’m an AI / Machine Learning Engineer focused on real-time Computer Vision, AI voice systems, cyber-physical AI, and applied machine learning. I build end-to-end ML pipelines, distributed backend architectures, and production-grade AI agents with PyTorch, scikit-learn, and TensorFlow.

Recently, I architected a real-time Voice AI platform using Python (LiveKit Agents SDK) and Node.js/TypeScript, enabling AI-driven outbound campaigns and customer support with <300ms end-to-end latency. I also reduced voice response latency by nearly 40% using a Speculative Race LLM approach, and integrated an STT → LLM → TTS pipeline with Deepgram Nova-3, OpenAI GPT-4o, and ElevenLabs—backed by circuit-breaker fallbacks to maintain 99.9% session uptime.

On the product side, I’ve built production-grade web interfaces using React, Next.js, and Tailwind CSS, supporting 100+ users and strengthening frontend-backend reliability through API testing. My projects include Smart Grid cyberattack detection using ST-GNNs, Transformer-LSTM hybrids, and a PPO-style mitigation agent (with high detection/localization performance), plus an AI civic cleanliness verification platform using CNNs/Vision Transformers, contrastive learning for before-after verification, and scalable SaaS design with MongoDB and REST APIs.