Know which satellite data and onboard-AI outputs you can trust.

An illustrative pass through a geomagnetic storm. Watch what Gauss does to each telemetry record and each onboard-AI output as conditions change.

1. T−24 minG0 · S0 · quietNominalRunInputs fresh, provenance complete.
2. T−12 minG1 · S0 · field turningNominalRunWatch: geomagnetic activity rising; no proton event yet.
3. Storm onsetG2 · S2 · protons risingDegradedDegradeProton flux up: bit errors climbing; outputs flagged, confidence lowered.
4. PeakG3 · S2 · stormBlockedAbstainPacket integrity failed in the proton event; no trusted input, so no output.
5. T+10 minG2 · S1 · easingUnknownFallbackGeomagnetic link gap: records stay unknown, never "nominal". Deterministic fallback used.
6. T+30 minG0 · S0 · recoveredNominalRunQuality gate passed; every flagged record is in the report.

Each step reads top to bottom: space environment (G = geomagnetic, S = solar radiation), then the telemetry record state, the onboard-AI output policy, and what the evidence records. Bit errors follow S; link gaps follow G.

Illustrative data, not a mission record. In a pilot, the replay runs on your own inference logs and telemetry exports, against the public space-weather record. Gauss is observe-only and advisory: decisions stay with your operators.

See a sample evidence report Start with a pilot

02_HOW_IT_WORKS

How a replay pilot works

Fleet Integration v0.1 runs on the ground, on your own logs. Nothing is installed on a spacecraft.

1. Bring your logs

   Onboard inference logs and telemetry exports (CSV, JSON or JSONL today; Yamcs and XTCE within the pilot scope), shared under a data-sharing agreement.

2. Correlate with the space environment

   Each record is lined up with the space-weather conditions at the time: radiation and geomagnetic activity from public NOAA and NASA HAPI services.

3. Score quality and provenance

   Every value carries its source and age. Missing or stale inputs stay unknown, never "nominal". What Gauss does with an output taken in a storm, we walk through in person, on your own logs.

4. Get the evidence

   A report your engineers and assurance reviewers can check, with success metrics agreed before the pilot starts.

Working today: live NOAA space-weather ingest, with NASA-archived storms for replay; three L1 spacecraft read, with a flag when SOLAR-1 and IMAP disagree on density; the solar wind propagated to Earth for about 50 minutes of warning; a Kp 3-hour forecast that beats persistence on held-out 2022–24 data, and is worse during storms, which we say here rather than hide; radiation (S), geomagnetic (G) and radio-blackout (R) levels kept apart, computed from GOES protons, Kp and GOES X-rays to NOAA’s thresholds and cross-checked against NOAA’s published scales; every input sense-checked, so inconsistent sensor data is flagged with a reason instead of dropped; stale or missing inputs shown as unknown (S?, G?, R?), never as quiet. Being built for the first pilots: ingest of your own logs, output flagging and the evidence export.

03_CLAIMS

What we claim, and what we don't

04_PLANS

Start with a pilot. Grow in gated stages.

1. Start here

   Design-partner pilot

   £20k–£40k per pilot

   One asset group, one historical replay, one live or near-live feed, one measurable acceptance outcome.

   • Replay of your inference logs and telemetry exports against the space environment they met
   • One live or near-live feed
   • One measurable acceptance outcome, agreed before day one
   • Provenance, freshness and uncertainty on every exported decision
   • Radiation and space-weather context correlated with telemetry quality
   • An evidence report and a review session with your team

   Bounded and paid: about 30 days, at a fixed price agreed from your scope before you pay.

   Start with a pilot

2. In development

   Fleet Integration

   £60k–£125k a year

   Recurring mission-data quality, provenance, radiation correlation and evidence exports.

   • Recurring mission-data quality checks across your fleet
   • Provenance, freshness and uncertainty on every exported decision
   • Radiation and space-weather context correlated with telemetry quality
   • Explicit nominal, degraded, abstain, unknown and blocked states
   • Evidence exports for engineers and assurance reviewers
   • Operator-visible incident and acceptance reports

   Usually follows a pilot. Billed yearly.

   Ask about Fleet Integration

3. In development

   Mission integration

   £120k–£300k one-off

   Your archive, mission dictionary, Yamcs and ground integration, and acceptance work.

   • Your archive and mission dictionary, integrated
   • Yamcs and ground-segment integration, with acceptance work
   • One measurable acceptance outcome, agreed before day one
   • Operator-visible incident and acceptance reports

   Scoped with your team, delivered by milestone.

   Ask about Mission integration

4. In development

   Flight Candidate Adapter

   £750k–£2.5m one-off

   Target-processor and mission-specific validation, after hardware and operator gates.

   • Target-processor and mission-specific validation
   • Operator-visible incident and acceptance reports

   Only with a named mission, target processor, mission dictionary, hardware evidence and operator authority.

   Ask about Flight Candidate Adapter

Your price is a fixed quote inside the range, agreed from your scope before you pay. Project work is billed by milestone. Every plan is advisory: Gauss shows evidence and never commands a spacecraft. Today we take on design-partner pilots; the later plans are in development and follow a pilot.

Later, a strategic option: Federated Bridge Node, a hosted-payload or in-orbit demonstration, considered only after the stages before it. Future work, not on sale.

05_FIT

Where Gauss fits

We design to standard boundaries. We don't replace your mission control.

• CCSDS and XTCE Packet identity and telemetry definitions, as your ground segment already uses them: the boundary a pilot integrates at.
• Yamcs (pilot integration) Designed to replay from mission-control archives; built with the first pilot.
• ONNX (pilot integration) For teams whose models run in orbit: scoped in the pilot, not before.
• Your operators decide Observe-only and advisory. Gauss consumes the mission data you already have and never replaces mission control.

Design-partner paths: we are working towards validation pilots with JAOPS and xOrbita.

06_CONNECTIVITY

Assured Connectivity for the team on the ground.

Argo keeps an operations team's laptops and ground hosts working on hotspots, hotel and café Wi-Fi. It finds what broke, fixes what it safely can, and tells a person what it can't.

• Self-healing Checks the connection every minute, names DNS and proxy faults and fixes the ones that need no administrator, opens Wi-Fi login pages, and remembers what worked on each network. One alert per problem, not one every few minutes.
• Power, memory and disk Watches all three, holds optional work back when they run short, and keeps a 10% battery floor.
• Wi-Fi security Flags weakly encrypted networks and devices left open to everyone on them. Advice only: it never changes your sharing or security settings.
• Exit country Checks which country your traffic leaves from and says when a SIM or network sends it somewhere unexpected.
• Team status Each device reports to the Gauss operator console, and a phone that drops off the team network gets a text with the fix.
• Mac and Linux Runs on macOS laptops, and on Linux ground hosts as a small non-root container.

Running today on Lightbound's own laptops. Offered to pilot teams, scoped per team.

Ask about a connectivity pilot

07_TEAM

Team

• Joshua Nathan CEO. Aerospace engineering (University of Sheffield), venture leadership.
• Shubham Hule CTO. Space-systems engineering (MSc, University of Southampton); working on a LEO satellite.
• Merve Nur Yaşar Chief Strategist. PhD researcher in astropolitics; space policy and regulation.
• Ashwin Sriramoju Applied AI and multi-agent systems; edge compute.
• Aatish Patel AI systems and revenue operations.