Insight-only. Built to last.
Custra exists because CRE portfolios need a coordination hub, not another point system. We built the hub, and we built it on a principle that doesn’t bend: observe, predict, advise — and never actuate.
The mission
Commercial buildings generate more fault signals than any facilities team can process. The result is reactive maintenance: tenant complaints, water damage reports, regulatory citations, and a backlog that grows faster than it closes. The problem isn’t data volume. It’s coordination: the right signal isn’t reaching the right person, interpreted correctly, in time to act.
Custra solves the coordination problem. One hub that reads every building system, maintains a live operating state, detects faults before the alarm fires, and routes a single correct response: one trade, one work guide, one verified close. The hub is passive by design — it never touches a thermostat, valve, or control system. The value is in what it knows and who it tells, not in what it controls.
What we believe
Causal beats correlational
Buildings are physically coupled. An HVAC overheat cascades to electrical load, CO⊂2;, and tenant comfort. Correlation finds the co-occurrence; causality finds the root. We built a graph learner that confirms causal edges interventionally — the only way to dispatch to the right trade instead of five wrong ones.
Density manufactures intelligence
One building can take years to learn its own coupling graph. 100 buildings learn it in 3 months. The federated layer is not a future product — it is the architecture decision that makes every building on the network smarter as the network grows. We designed for density before we had the density.
Insight-only is a commitment, not a phase
Every IoT platform that starts as “insight-only” eventually adds actuators — because actuation is where the next dollar looks like it lives. We disagree. The coordination value is permanent; the actuator temptation is a governance risk. Custra never actuates. Not in this version, not in the next one.
The evidence gate is non-negotiable
We built on synthetic data and proved the case lifecycle end-to-end before seeking a single design partner. The next gate is real: detection must be proven on real building fault data before we claim it works on real buildings. We are honest about where we are in that sequence.
Powered by Atlas
Custra is a product of the Atlas program — a family of industrial intelligence platforms built on a shared, composable spine. The Atlas family includes:
- Atlas IS — Infrastructure state: fault detection and advisory for physical building systems.
- Atlas MO — Movement and dispatch: service coordination, escalation lifecycle, and vendor sourcing.
- Atlas OI — Operational intelligence: multivariate anomaly detection and advisory for building operations.
Custra composes all three engines and adds the coordination hub, portfolio layer, federated intelligence, and the Atlas Inside network. The Atlas composable coupler (atlas_compose) fans every pipeline stage across all three engines and merges outputs namespaced per engine — the same spine that powers every Atlas family product.
Every new Custra capability is a plug-in on the Atlas spine. Every new vertical is a configuration, not a codebase fork. The architecture was built to be composed, extended, and federated from the first commit.
Atlas Inside
Custra is the first product on the Atlas Inside network — the two-sided platform that connects building owners with service providers and turns anonymized cross-building learnings into a shared intelligence asset.
The network grows in value as it grows in buildings. Every owner added makes the coupling graph converge faster for every other owner. Every provider on the network raises the quality of dispatch for every owner. The flywheel turns when the evidence gate is cleared and the first design partners join the network.
We are building toward that first turn now. Contact us to join as a founding network member.
Where we are
Design-partner phase. The platform is synthetically proven: 24 building systems, 103 sensor families, and the full service-case lifecycle — detect, triage, authorize, dispatch, repair, verify, bill, close — running end-to-end with precision and recall of 1.0 across edge cases including compound fault stress. The causal graph learner has been validated on Monte Carlo simulations of 10 to 10,000-building fleets. The next step is real building data, and for that we need design partners — portfolio owners, REITs, and facility managers willing to bring real fault logs and help us prove detection before we go to market at scale.
This is a direct, honest account of where Custra is in 2026. The design-partner engagement is structured to deliver value to the partner on real data from the first engagement — not to extract data as a free trial. We work through the evidence-gate proof together, and the partner gets a building-intelligence capability that is proven, not promised.
Contact
For design-partner inquiries, technical questions, or any other conversation:
Richard Bayley
richard@accelerate-ip.co
831-431-7147