Case Study: NASA improves space communications and mission workflows with Deepgram

NASA uses Deepgram to power the next generation of space tech

NASA partnered with Deepgram to improve speech-to-text for its noisy, hard-to-understand audio from space and training environments. The agency needed fast, accurate automatic transcription for workflows like space-to-ground communications, historical mission search, and future astronaut support, but existing speech-to-text tools and an in-house open-source solution did not reach the 80% word recognition rate NASA required.

Deepgram provided its AI Speech Platform, including tailored speech models and audio search, to train on NASA-specific audio such as space-to-ground communications and Neutral Buoyancy Lab recordings. The result was a custom model that reached up to 89.6% accuracy on space-to-ground transcripts and about 87% word recognition on NASA validation sets, enabling new use cases like detecting readback errors, searching mission archives faster, and powering the IRIS medical chatbot concept.

NASA

Clinton May

Federal Strategist