"Hey Diego" — a crewmate, not a console.

Today, dozens to hundreds of flight controllers and subject-matter experts run real-time troubleshooting, just-in-time training, and procedures from the ground. Beyond low Earth orbit, that lifeline thins. Crews need to interact with a spacecraft's autonomous systems directly — hands-free and eyes-free, because in space both hands are usually busy.

The system had to do more than execute commands. On a long-duration mission, a voice in the habitat is part of the crew's psychological environment.

Long-duration spaceflight brings motivational decline, fatigue, insomnia, and social tension. Astronauts surrounded by peers can still feel lonely. The brief we set ourselves: a voice management system that supports daily operations and behavioral health — raising alerts in a calm voice rather than with alarms.

Voice assistants on station have a history — Clarissa (NASA/Xerox, demoed 2005 at Michigan), Airbus's CIMON on the ISS, and NASA Stennis's AVA prototype. CIMON's lessons loomed large: it wasn't fully integrated into the ISS, and it wasn't accessible everywhere. And existing voice technologies wear thin fast — they misinterpret you, then make you repeat yourself.

Each Voice Design team member drafted a complete human persona — biography, voice, temperament — and we pressure-tested them against the survey traits. The result was Diego: kind yet logical, expressing personality through vocabulary, varied phrasing, and humor in regular dialogue.

Every component had to run offline — no internet beyond LEO — on modest hardware: development on Ubuntu, migrating to two Raspberry Pis for integration testing. Trade studies selected each stage: Porcupine for wake-word detection ("Hey Diego" is the only phrase that wakes it), Mozilla DeepSpeech for transcription, Rasa for intent and entity extraction, and Coqui for a natural text-to-speech voice.

Unlike CIMON, Diego is built into the station itself: fixed microphones in every module, and a bridge into NASA's Platform for Autonomous Systems (NPAS) so a spoken request becomes an executable command set on the spacecraft's own autonomy layer.

A station is loud, and a voice interface that keeps mishearing people will simply get turned off. Scott Kelly wrote about feeling measurably "stupid" as CO₂ rose; a system that monitors vitals by voice has to work in the same harsh acoustic environment it reports on. The Background Noise Mitigation team attacked the problem from both ends: passive cancellation (foam panels absorbing high frequencies) and active cancellation (anti-noise for low frequencies).

The project delivered a working prototype interface compatible with NPAS, software demonstrating intent interpretation and appropriate response, expanded conversations across all four request categories, and hardware/software recommendations for noisy environments — plus an Adobe XD GUI (home screen, calls, messaging, notes, scheduling carousel, vitals charts) for moments when voice isn't the right channel. Published with our NASA Stennis collaborators.

The surveys turned warmth into a real requirement. On a long mission, the crew hears this voice more than they hear most of each other — every day, for months. A system can be accurate and still wear on the people who have to live with it.