I like working with open-source natural language models and vision models, especially around data, eval, and post-training. I’m into agentic systems and specializing models for specific domains, and I enjoy turning research ideas into things people can actually run, modify, and build on.