Worked on a peronsal project to use AI to help determine where in a long form youtube video a certain topic is discussed. User inputs a youtube video link and a topic, a simple frontend webpage calls a backend service written in Python and returns the youtube link with a timestamp parameter that allows user to skip to the point in the video where the desired video is discussed.

Backend service uses Pinecone, which is a vector embedding database, to upload vector embeddings of the video transcript. Pinecone is then used to determine where in the video the desired topic is discussed by seeing which vector embeddings are closest to the topic query vector embedding.