Welcome to Yingtao’s Homepage!

Biography

My name is Yingtao Luo (Ying-Tao Luo, 罗颖韬). I am a final-year Ph.D. candidate in Machine Learning (joint program) at Carnegie Mellon University, affiliated with Heinz College and Machine Learning Department.

My research broadly focuses on LLM post-training, generative retrieval, personal agents, long-horizon sequence learning, and reinforcement learning.

My advisory committee members are Prof. Rema Padman at CMU, Prof. Reza Skandari at Imperial College London, Prof. Bryan Wilder at CMU, and Prof. Arman Kilic at MUSC. Together, we work on an event-driven offline RL framework for irregular, long-horizon decision-making with delayed feedback and severe action imbalance. Our approach combines Transformer-based state modeling, conservative policy learning, and off-policy evaluation, with LLM agents serving as a conversational interface between the learned decision policies and human decision-makers. I also work extensively on personal agent research and production-grade personalized generative retrieval systems.

I have previously worked as a research/applied scientist intern at RealAI, Microsoft, Alibaba, and Roblox, with approximately two years of industry experience in total.

I am deeply grateful to my mentors, collaborators, and friends, from whom I have learned tremendously. Their support has played an important role in shaping both my research and who I am today.

Always happy to chat about jobs, startups, research, or interesting ideas!

Research Interest

LLM Post-Training · Generative Retrieval · Personal Agents · Long-Horizon Sequence Learning · Reinforcement Learning