It required a combination of edge computing, real-time streaming architecture, and model-level optimization. We decoupled our speech-to-text and …
http://dlvr.it/TLM2s5