Summary
Puyuan Peng is a research scientist focused on SpeechLLMs, currently at Meta’s Superintelligence Lab after completing a PhD at UT Austin and seven years of research experience. His work spans zero-shot TTS and speech editing (VoiceCraft, ACL2024) and audio-visual generative models (ECCV2024), with VoiceCraft notably trending #1 on GitHub and amassing 8.4k stars. He pioneered prompt-based techniques for large speech models (PromptingWhisper) and has driven state-of-the-art results in visually grounded speech tasks, zero-resource recognition, and data-efficient representations. Puyuan blends strong statistical and mathematical foundations (MS Statistics, BS Mathematics) with hands-on experience training SpeechLLM pretrained models and audio tokenizers, and he has a track record of translating cognitive-inspired insights into scalable ML systems. An underappreciated strength is his consistent ability to produce transferable, zero-shot capabilities that bridge research demos and real-world VR and multimodal applications.
7 years of coding experience
4 years of employment as a software developer
Bachelor of Science - BS, Mathematics, Bachelor of Science - BS, Mathematics at Beijing Normal University
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at The University of Texas at Austin
Master of Science - MS, Statistics, Master of Science - MS, Statistics at University of Chicago