Story

In brief

We are investigating whether recurring LLM workloads can be replaced, where appropriate, by automatically constructed pipelines of regexes, deterministic parsers, traditional ML and NLP models. As an example, suppose...