Ground segment · Autonomous AI
VELA turns a telemetry deviation into a cited root-cause narrative and a drafted recovery procedure — reasoning across the spacecraft's own manuals and knowledge graph1 and showing every step of its work. Runs fully air-gapped.
Act 01 — Detect
A channel breaks its own rhythm. Onboard FDIR flags it, or VELA's own detectors catch it over the downlink. Before anything else, the telecommand-response filter removes deviations that simply followed a command2 — so the investigation starts on a real fault, not on noise.
ECOD · graph · onboard-AI · seeded replayAct 02 — Explain
A deterministic agent graph walks the anomaly through the spacecraft's knowledge graph, reaching data only across an audited boundary. It writes a root-cause narrative where every claim resolves to a manual section, an FMECA entry, or a graph fact — or it is blocked as UNGROUNDED and never shown as fact.
Battery cell degradation — string 3 undervoltage
Deviation follows no related telecommand; the cell-imbalance signature matches the FMECA degradation mode for the EPS battery.
Act 03 — Resolve
VELA drafts the recovery procedure the anomaly calls for — every step validated against the spacecraft's manuals and cited to its source, wrapped in a full audit trail. Your team gets a ready-to-run plan in minutes, not a blank incident doc and a war room.
validated procedure · append-only auditGrounded, not guessed
Most AI asks you to trust a black box. VELA reasons across the spacecraft's own manuals, FMECA, and knowledge graph and cites every claim it makes — or it says, plainly, that it can't. The traversal, the citation checks, the replay: deterministic and verifiable. That is what makes autonomous AI usable on a real fleet.
Why teams trust it on a live fleet
Six properties that make an AI safe to put in front of a spacecraft operator.
Every narrative and procedure claim resolves to the spacecraft's own manuals and graph — or it is emitted as UNGROUNDED and held from release. A hallucinated reference is a sev-1.
From the first deviation to a drafted recovery, VELA runs the whole investigation itself — triage, hypothesis, validation, and drafting — as a deterministic agent graph that reaches data only through an audited boundary.
Runs fully offline with in-memory stores, or against your own Neo4j and Qdrant. No external calls, no CDN, no data leaving the ground segment.
Telemetry enters through one swappable source — CCSDS / PUS, XTCE, Yamcs, OpenC3, MQTT, or seeded replay. The reasoning above it never changes.
Graph traversal, event forming, and citation checks are deterministic and unit-tested; replay is seeded, so an investigation reproduces exactly — every time.
Beyond live faults, VELA projects remaining-useful-life from each component's degradation trend — surfacing the failure that is coming, not only the one that already happened.
Plug into every tier — one swappable telemetry source
See it investigate
A guided walkthrough on the VELA-REF-1 reference spacecraft — no live mission data required. Run it in your own air-gapped environment.
VELA