Projects

  • Learning NLP through steam game reviews

    There are over 100,000 games in the Steam catalog and 132 million monthly active users (MAUs). Users are allowed to leave reviews and votes (thumbs up or thumbs down) about games after purchasing.

  • Chicago wildlife species analysis - a starting point

    After improving the species classification model to a point where we can start to trust its predictions - we ran it on 71k+ images where speciesnet predicted at least 5% likely containing an animal.

  • Filtering 90% of blank wildife photos for a better UX

    Our community wildlife classification/identification project is live at rangers.urbanrivers.org.
    Using AI, we are massively fixing the boring part.

  • Comparing AI computer vision with Human Labeling

    We have AI results and volunteer contributed human labels for approximately 60,000 images.
    This post explores the relationships between human error and ai errors.