One 12nm 18.1TFLOPs/W Transformer accelerator (led by Harvard) and a 21.38TOPS/W compute-in-memory macro design (collaborated with Prof. Le Ye).