Startup Builds AI System To Transcribe Meetings

Voicera, a San Francisco Bay Area startup, recently announced $20 million funding for their GPU-based deep learning system that can now fully transcribe meetings and put together highlights.

The system was designed to help teams better collaborate in an enterprise environment.

Eva, the start-ups AI assistant, joins meetings and conference calls through a combination of machine learning, voice recognition, and natural language processing. The service provides customers with meeting recordings, transcriptions, and highlights, which can then be shared and edited through collaboration systems like Slack and Salesforce, as well as via email.

The team designed a system that can handle multiple speakers and deliver higher accuracy rates than current subscription services.

To develop the system, the team used their local cluster of eight Quadro P4000 GPUs with the Torch deep learning framework to train their neural networks. Once trained, they use GPUs on the Google Cloud with TensorFlow to handle the inference of their neural nets.

“We believe that voice technology can transform how we work and a huge portion of our time at work is in meetings,” said Voicera CEO, Omar Tawakol in a company release. “We wanted to make meetings actionable,” he explained.  

The system is built to work in meeting environments such as BlueJeans, Zoom, UberConference, Cisco WebEx and Skype.