Haosong L.

I’m an Applied AI Engineer specializing in building real-time AI agents, including voice and multimodal systems, with a strong focus on low-latency performance and production reliability. At Newcast AI, I design and deploy end-to-end solutions using LLMs, RAG pipelines, and streaming architectures (Pipecat, LangChain, Qdrant), achieving sub-second response times in live environments. I have hands-on experience in prompt engineering, conversation design, and system optimization, creating structured, stateful AI workflows that improve efficiency and user experience. My work includes deploying scalable AI services such as voice cloning and lip-sync models, as well as developing intelligent customer support agents with strict grounding and step-by-step reasoning. With a background in data science and machine learning, I bridge the gap between research and real-world applications—turning cutting-edge AI into reliable, user-facing products.