Agenda

I would like to present how to use weak supervision, LLM's, and embeddings for Discriminative AI like approving a loan or a medical procedure pre-certification. Typically we use statistical methods for such classification tasks working on a structured dataset. In this demo, I will train a classifier that will make decisions from unstructured text (Doctor/Clinical PA notes, Plan Benefit document etc). We bring in reasoning/embeddings from LLM's and use it as a labeling function along with other heuristics/dictionaries for classification. Data-Centric AI.