Deepgram co-founders, 20-year-old wunderkind Noah Shutty and his University of Michigan lab supervisor Scott Stephenson, definitely have faith in sound.
The two left behind promising research in particle physics, researching the creation of Dark Matter in the bowels of underground Chinese bunkers, to develop what they call a “Google for sound”.
Much as the Google search engine uses links between different web sites to surface the most relevant search results, the Deepgram software uses connections between certain phrases in audio files to identify the type files and yield relevant results.
This audio search technology is much different from others that rely on natural language processing, according to the founders. “We do train models to tell the difference between calls,” says Stephenson in an interview. “We’re doing it in terms of context.”
The business started as a side project for Shutty — a way for him to organize his thousands of hours of “lifelogs” — the audio and video he’d amassed recording every aspect of his life.
He wanted a way to search through his audio files and developed a neural-net-based artificial intelligence.