Huck Yang

About

I am interested in scaling (i.e., train; verification), text-space alignment of audio, computer-apps and voice-agents. I worked at NVIDIA Research, Amazon ASR-LM, working with Andreas Stolcke, and as a research intern at Google (now DeepMind), co-hosted by Bo Li; Yu Zhang in Tara N. Sainath's team.

🎓 My thesis was on privacy-preserving post-training, advised by Prof. Chin-Hui Lee.

🧬 I visited Prof. Jesper Tegnér's group on self-evolutionary ML, interned at TSMC in RL for mixed-signal IC design before Georgia Tech.

• Multimodal Post-Training: Co-led long omni / audio traces for Nemotron-Omni-30B [GTC 26], and long-horizon alignment in OmniVinci-9B [ICLR 26], including RL post-training on audio [ICLR 25, 24]. Google-affiliated patent on multilingual ASR [ICASSP 23]; best-paper nominee at [Interspeech 23]. The first voice prompting method to frozen acoustic models [ICML 21].
• Voice Interative Agents: the first series of n-best hypotheses based generative ASR correction (GER) in [ASRU 23] and Whispering-LLaMA [EMNLP 23]; HyPoradise in [NeurIPS 23], and audio post-training [ICLR 24]. Best industry paper honoronable mentioned award 🎖️ on multimodal correction in [ACL 25]. Self-reflection agent: Speech-Hands [ACL 26].

⚛️ Fun fact: I also work on Quantum ML part-time for fun, where I created the first variational circuit based speech [ICASSP 21] and text classification [ICASSP 22] and received the Xanadu AI Quantum ML Award in 2019; recently, on quantum parameter adaptation for LLMs in [ICLR 25].

Tutorials

Blog