Can AI design circuit boards from text descriptions?

PCB schematic design is tedious and manual. SchGen, a specialized LLM, generates editable schematics from plain-language requests by using a semantically grounded code representation that strips away geometry-heavy details and focuses on wiring logic. The team built a large paired dataset via human-AI collaboration, converting open-source hardware into their representation. On wire connectivity and functional correctness, SchGen outperforms both naive representation schemes and larger general-purpose LLMs.