Deepgram aims to run voice AI natively on air-gapped laptops
Deepgram staff technical program manager Nick Mann told theCUBE at Dell's "AI Leadership Symposium" that the company wants its voice models to run natively on laptops and eventually work air-gapped on-device. He noted Deepgram expanded its Flux model to 10 languages earlier this year, that its 45 researchers consume large amounts of compute and data, and that a large share of customers host the models themselves, with a Fortanix partnership bringing private, on-premises voice AI to regulated industries. Mann said Dell supplies Deepgram's infrastructure and also runs its models in Dell's Customer Solution Center Innovation Lab, and that some small businesses already run workloads such as interactive voice response on a compact Nvidia GB10 system.
On-device, air-gapped voice AI bears on privacy and latency, and shapes whether voice model vendors can serve schools, emergency services and other data-sensitive settings.