NASA tests local LLM inference for future space missions

Red Hat and NASA researchers are developing the Crew Medical Officer Digital Assistant (CMO-DA), a medical AI system that runs large language models on local hardware with zero cloud dependency. This initiative addresses the impracticality of Earth-based telehealth for astronauts on Moon or Mars missions due to light delay and communication blackouts.

The CMO-DA utilizes RamaLama, an open-source CLI tool backed by Red Hat, to manage inference via llama.cpp and other engines like MLX and vLLM. The system treats AI models as portable artifacts, enabling reproducible and cryptographically verifiable deployments on edge hardware. NASA is currently testing the system on terrestrial hardware that mirrors the HPE Spaceborne Computer aboard the International Space Station.

NASA adopted a local-first architecture because onboard compute constraints and the need for reliable medical decision-making outputs leave no other viable option. This deployment demonstrates the critical role of local and open LLMs in supporting future deep space missions.