I am a Staff Machine Learning Engineer with over five years of experience building production-grade AI systems across deep learning, computer vision, generative AI, agentic AI, and multimodal applications.
I specialize in designing end-to-end real-time AI solutions, including Video AI, action recognition, voice-to-voice agents, RAG and GraphRAG pipelines, and LLM-powered enterprise workflows. My work emphasizes low-latency inference, scalable orchestration, structured outputs, and reliable production deployment.
I have built multi-provider voice AI platforms integrating Gemini Live, OpenAI Realtime, Qwen Omni, WebSockets, telephony systems, LangGraph, Redis, and FastAPI. I also develop resilient AI systems with function calling, session management, prompt templating, provider failover, and context-aware enterprise retrieval.
My computer vision background includes sports video analytics, highlight generation, object detection, OCR, action recognition, replay detection, segmentation, pose estimation, and logo tracking. I have worked extensively with PyTorch, YOLO, VideoMAE, CNN-RNN architectures, attention mechanisms, Whisper, and multimodal foundation models.
I am experienced in AI safety and evaluation practices, including hallucination detection, evidence verification, LLM-as-a-judge evaluation, prompt-injection defense, Pydantic validation, and JSON Schema-based structured outputs. I focus on translating complex AI challenges into measurable improvements in accuracy, automation, latency, and operational efficiency.
I hold a Master's degree in Artificial Intelligence and Machine Learning and have contributed to research in self-supervised contrastive learning, video understanding, action recognition, machine learning, and gene variant classification.
