Daphne Cornelisse
Daphne Cornelisse
Embedded Files
Hello! I am a third-year PhD student at NYU, advised by Prof. Eugene Vinitsky. I build fast multi-agent simulation environments and explore how to train effective agents by combining self-play RL with human data. Before my PhD, I completed an undergraduate degree in Neuroscience and a master’s in AI, where I worked on cooperative game theory.
My work is supported by a CAIF Fellowship and a Chishiki AI Fellowship (UT-Austin and NSF).
I interned at Waymo in the summer of 2025.
News
Dec 2025. Release: We released PufferDrive 2.0! Train driving agents from scratch in 15 minutes and drive with them, too.
Oct 2025. Blog: How to catch subtle RL bugs before they catch you.
Oct 2025. Blog: Human behavior modeling in naturalistic driving: Trends and opportunities.
Jul 2025. Talk: Gave a talk at Waymo on guided self-play.
May-Aug. Internship: Spent the summer interning at Waymo's safety research team.
Apr 2025. Fellowship: I was selected as a Chishiki AI Fellow (funded by NSF).
Mar 2025. Talk: Gave an invited talk at the AI and Social Good Lab (AISOC) at CMU.
Jan 2025. Workshop: I am co-organizing the workshop on Computational Models of Human Road User Behavior for Autonomous Vehicles at IEEE IAVVC.
Jan 2025. Fellowship: I'm honored to have been awarded the 2025 PhD Cooperative AI Fellowship!
Oct 2024. Talk: Gave an invited talk on HR-PPO and GPUDrive at the RL reading group at UoE! You can watch the recording here.
May 2024. Paper: Human-compatible driving partners through data-regularized self-play reinforcement learning was accepted to RLC 2024.
Apr 2024. Talk: Gave an invited talk on Human-Regularized PPO at the Berkeley Multiagent Learning Seminar! The slides are here.
Sep 2023. Started my Ph.D. at NYU with the EMERGE Lab!
Selected projects
PufferDrive 2.0: A fast and friendly driving simulator for training and evaluating RL agents
Daphne Cornelisse*, Spencer Cheng*, Pragnay Mandavilli, Julian Hunt, Kevin Joseph, Waël Doulazmi, Valentin Charraut, Aditya Gupta, Joseph Suarez, Eugene Vinitsky
Software release
GitHub | Tweet | Release post
Building reliable sim driving agents by scaling self-play
Daphne Cornelisse, Aarav Pandya, Kevin Joseph, Joseph Suarez, Eugene Vinitsky
In submission
Paper | Tweet | Project page
Neural payoff machines: predicting fair and stable payoff allocations among team members
Daphne Cornelisse, Thomas Rood, Mateusz Malinowski, Yoram Bachrach, Tal Kachman
NeurIPS 2022
Paper | Poster | Tweet | Master thesis
Page updated
Google Sites
Report abuse