ChipNeMo, trained on Nvidia’s internal repositories of code and text, can answer questions about chip design, summarize bugs and write scripts for EDA tools.