Patent-pending inference architecture — technology that connects dots on its own.
U.S. provisional filed — 50 claims, three families
GeoFiber is an inference engine that finds what’s wrong before you know to look for it. It learns what normal looks like by comparing every entity to its peers, flags the one drifting away, and tells you exactly which dimensions changed and why — no black-box scores.
It was built on a conviction: the relationships between things — patient to cohort, machine to fleet, host to network, instrument to market — carry more signal than the things themselves. Inference that respects that structure finds what rule-based and score-based systems cannot, and can explain what it finds.
No thresholds, no hand-written rules. GeoFiber learns what normal looks like for each entity from the behavior of its peers — and keeps learning as the population moves.
Every detection names the dimensions that changed and the peer context that makes the change anomalous. Operators get evidence they can act on; reviewers get reasoning they can audit.
When an entity falls outside what the engine can support, it says so — a calibrated statement of confidence instead of a confident guess. The property assurance cases are built on.
The same architecture reads patients, machines, hosts, spacecraft, and instruments. Retargeting is an ontology swap, not a rebuild — one engine, many missions.
Fleets of nominally identical aircraft, spacecraft, and vehicles never behave identically. GeoFiber learns each asset's expected envelope from its peers and surfaces the tail number drifting from the fleet — deviating dimensions named, early enough to act. Applicable to integrated system health management, test & evaluation telemetry, and sustainment analytics.
Autonomy that can't explain itself can't be accredited. GeoFiber's outputs carry their reasoning and a calibrated confidence statement, and the engine declares when it is operating outside what it knows — properties aligned with explainable-AI and mission-assurance requirements.
High-dimensional observational data hides structure that summary statistics average away. GeoFiber surfaces the entities, cohorts, and interactions that don't fit the population — a systematic instrument for anomaly-driven discovery in experimental and observational datasets.
The commercial deployments of the same engine: clinical and revenue operations, industrial reliability, security operations, and market intelligence. Plug in your data; curvature finds what your dashboards can't.
GeoFiber is dual-use by construction: the same engine that audits claims data can baseline a fleet, a network, or a constellation. Build TBD is actively pursuing research collaborations and federally funded R&D — SBIR/STTR programs, BAA responses, and teaming arrangements with primes and research institutions — where explainable, peer-contextual inference addresses program goals in system health, assured autonomy, and anomaly-driven discovery.
Status: U.S. provisional patent application filed (50 claims across three families). Working implementations demonstrated across domains. Technical discussions available under NDA; capability-level briefings available to program and proposal audiences without one.
The fabric on our homepage is this idea, animated — see it move at buildtbd.com
