New case study on training speech-to-text for restaurant industry. (Github included) 45% WER improvement, 360ms latency https://www.latice.ai/case-study What do you guy's think about that ?