A collection of non-research projects I've worked on over the years.

I also tweet and blog about my projects!

Some (current) personal favorites:

  • Clinical AI:

    • MedQA: a GPT-powered clinical reference tool that can answer clinical questions and follow-up questions in natural language, along with references to sources.
    • clinitokenizer and clinisift, BERT-based off-the-shelf tools for medical sentence tokenization and Named Entity Recognition, respectively
    • Clinical GPT-3, an exploration of OpenAI GPT-3's abilities to perform various Natural Language Processing tasks in the clinical domain, on patient medical records.
  • AI Software:

    • Museum Music, an app to use in museums, that dynamically updates a Spotify playlist with period-appropriate music by taking a photo of the art you’re viewing! Built with GPT-4-Vision and Spotify API.
    • Memcast, capture interesting parts of a podcast with just a screenshot! If you hear something interesting on a pod, but don't want to stop to write it down, just take a screenshot! Memcast will use it to auto-identify pod, episode, and timestamp.You'll get summarized bullet points from that timestamp, logged to your memcast notebook πŸ“’. Built with Whisper, GPT-4-Vision, GPT-4.
  • Emacs:

    • emacs-secondmate, a "mini-imitation" of GitHub Copilot for Emacs via HuggingFace Transformers
    • gpt-emacs-macro, running "semantic macros" in Emacs via GPT-3 (and a blog post on Semantic Macros in Emacs)