Building speech-native AI systems that can listen, reason, act, remember, and learn.
Currently, I focus on:
Speech LLM · ASR · Reinforcement Learning · Voice Agent · Retrieval · Memory
Sequence-level optimization for ASR and Speech LLMs, especially objectives beyond token-level cross entropy such as WER, CER, retrieval accuracy, and task-level rewards.
Speech to speech agent, contextual biasing, hotword retrieval, personalized memory, and retrieval over long-term spoken interactions.