Hi, I'm Kevin! I am a PhD student at BAIR under Pieter Abbeel and Sergey Levine. Before that I did my undergrad + master's at MIT CSAIL, under Phillip Isola.

My interests are in agent learning and AI-based creative tools. Some questions: how do we build agents that benefit from task diversity? Is there a data-heavy view of reinforcement learning that makes more sense? What tools will let humans be 10x more powerful in producing creative work?

In the past, I spent a year as a researcher at Cross Labs investigating open-ended learning. At Sizigi, I built AI-powered artistic tools. Before that, I worked on generative models at Autodesk Research and was on the reinforcement learning team at OpenAI.

My guilty pleasure is game design. I've released a few titles onto Steam: RAIN Project, Iwate Mountain Dance, Skyseeker. I am antagonistically interested in the psychology behind what makes games engaging. I am also deeply curious about procedural generation, and how AI/evolution/open-endedness can help us build virtual worlds from the ground up.

If you are interested in the topics above, shoot me an email (hi at kvfrans.com)! It helps if you've been brainstorming a specific idea or started some preliminary work; we'll have a starting point to discuss.

Timeline

  • 2017: OpenAI
  • 2018: Autodesk Research
  • 2018: Start undergrad at MIT
  • 2019: Sizigi
  • 2020-2021: Gap Year @ CrossLabs
  • 2021: Started ParagraphAI
  • 2018-2023: Phillip Isola's lab at MIT CSAIL
  • 2023-Now: BAIR