Dawen Liang

Email: email address

I am a machine learning researcher at OpenAI. Previously I worked at xAI and Netflix, focusing on LLM post-training, policy learning, and personalization. I completed my Ph.D. in the Electrical Engineering Department at Columbia University, as part of the LabROSA, working with Professor Dan Ellis and Professor David Blei. Prior to that, I completed my master's degree in Music and Technology program at Carnegie Mellon University, working with Professor Roger Dannenberg in the School of Computer Science, where I was part of the Computer Music Group.

Publications

For the most up-to-date information, please see my Google Scholar page.

Miscellaneous

Here are some old notes on topics I once spent time learning about. They unfortunately become increasingly irrelevant in the current age of LLMs, but I keep them here as a small piece of technical nostalgia. I hope they may still be useful to someone, though they may contain mistakes or outdated information. If you notice any problems, please let me know.

A proof that Maximum Likelihood Estimation (MLE) is biased for variance estimator.
A note that shows some useful properties for matrix derivative.
Some technical details for the Expectation Maximization (EM) algorithm.