Science-based · AI models · Live inference

Buildings lose energy through faults
that no single dataset
can see on its own.

When the board asks about energy, when an audit is closing in, when residents won't stop calling about cold apartments — Struxiva gives you the answer in minutes. From the documents and sensor data you already own, building by building, fault by fault, metric by metric.

In Finland alone, ~850 property management companies oversee 50,000 residential buildings and more than €8.5 billion in annual operating and renovation spend — without portfolio-wide visibility into where energy leaks.

Floor plan — fault detection active
ZONE A ZONE B ZONE C ZONE D ZONE E FAULT 01 FAULT 02
SENSOR ACTIVE
FAULT DETECTED

Buildings have data.
What they don't have is a dataset.

Around every building sits decades of drawings, reports, certificates, meter logs, sensor streams, work orders, and emails. The data exists — it's just scattered across tools, out of sequence, disconnected from its context, never validated against itself, and never assembled into something a machine — or a human — can reason over.

What operators actually wake up worrying about
  • The board is going to ask me again why the energy bill went up — and I still won't have an answer.
  • The audit is in six months, and I'm not ready.
  • Residents won't stop complaining about cold apartments in Block C.
Five gaps · one layer that closes them
Today
Not centralized. Drawings in one cloud, BMS in another, certificates in email, work orders in a spreadsheet.
Struxiva
One unified index for every file, feed, and record tied to a building.
Today
Not time-aligned. Documents span 30 years of revisions. Meter data streams every second. Nothing shares a clock.
Struxiva
Every record — static or live — placed on one canonical chronological axis.
Today
Not connected. A room in a floor plan, a sensor in a BMS, a setpoint in a report — same physical thing, no link between them.
Struxiva
Graph-level linking across documents, sensors, and systems — down to the individual valve.
Today
Not validated. Extracted values contradict each other. Sensor readings drift. No one knows which number to trust.
Struxiva
Cross-source validation with an explicit confidence score on every element.
Today
Not a dataset. Raw PDFs and CSVs can't answer a query, feed a model, or pass an audit.
Struxiva
A machine-readable, BRICK-native dataset — ready for analytics, audits, and AI.

Five steps from raw building data
to an actionable fault report.

The Struxiva pipeline is fully automated. Feed it documents and sensor feeds; receive a prioritised list of what's wrong, where, and what fixing it is worth.

01
Ingest & extract
Drawings, BIM models, technical PDFs, energy certificates, work orders, BMS feeds — regardless of format or age. Rooms, zones, equipment, and their relationships are extracted into one machine-readable model of the building's design intent.
Documents + sensors
02
Cross-validate against measured reality
Every extracted element is checked for internal contradictions, then compared against live operational time series — energy, CO₂, temperature, indoor air. Confidence score on every element; nothing reaches you that wasn't reconciled with a real meter.
Validation
03
Map to BRICK & validate the graph
The validated dataset is mapped to the BRICK ontology — a standard schema for building systems — and checked against SHACL (Shape Constraints Language) rules. Schema compliance and graph-level integrity are enforced before any fault analysis runs, so what you see is portable, audit-ready, and machine-verifiable.
Schema compliance
04
Detect faults & inefficiencies
Detects mismatches between design intent and operational reality — overventilated zones, stuck dampers, unoccupied-hours heating, setpoint drift, and other common but invisible fault classes.
Detection
05
Deliver an actionable, BRICK-native output
Each finding ships with fault type, precise location, recommended action, and a euro-denominated savings figure — portable, audit-ready, and ready for any analytics, audit, or AI tool you already use.
Output
Pilot finding  ·  1984 residential block  ·  4,820 m²  ·  Helsinki metro  ·  operator anonymised 1 of 6 findings shown
Issue
Ventilation running at full occupancy schedule during 22:00–06:00 — unnoticed since 2022 commissioning.
Location
Zone E · Floor 2
AHU-03 supply branch
Recommended action
Align AHU-03 schedule with occupancy sensor data. Reset setback from 21°C to 18°C for unoccupied hours. ~30 min on-site work.
Annual savings (verified)
8,400 kWh
≈ €1,050/yr · payback < 1 week · operator confirmed ±18%
Or ask any metric — SRI · EUI kWh/m² · PMV/PPD · U-values · Air-change rate · Thermal bridges · CO&sub2; compliance · District-heat €/m² · or your own KPI

Built on peer-reviewed science,
not marketing math.

Struxiva's fault detection, savings estimates, and ontology layer are grounded in published research and open standards — not vendor benchmarks.

01
LBNL 550-building study
Our savings estimates are derived from Lawrence Berkeley National Laboratory's landmark fault-detection study across 550 commercial and residential buildings — the largest validated dataset in the field.
02
BRICK ontology
We map every building to BRICK, the open W3C-compatible schema co-developed by UC Berkeley, EPFL, and industry partners. Your data is portable, auditable, and machine-verifiable.
03
ASHRAE & EPBD alignment
Fault classes and energy performance calculations follow ASHRAE Guideline 36 and EU Energy Performance of Buildings Directive thresholds — the same standards your auditors use.

Four kinds of teams — one platform
that speaks to each of their needs.

Struxiva adapts to the decision-maker, not the building type. Whether you manage one property or a thousand, the same pipeline delivers the answers that matter to your role.

Portfolio managers
You answer to dozens of boards. Always-on fault reports for every building you manage, with cost estimates ready to present at the next meeting.
Property owners & boards
Rising bills, no technical team. Plain-language reports on what's wrong, where, and what fixing it is worth — make the next renovation decision with evidence.
Public & institutional real estate
Mandates, audits, public capex. One audit-ready, BRICK-native dataset across your entire portfolio — standardised, traceable, and defensible on day one.
Facility service providers
Reactive maintenance is a losing game. A predictive intelligence layer you resell — spot faults before your clients do and turn maintenance into a premium service.

Built for how buyers actually evaluate
building intelligence — not how vendors pitch it.

These are the four reasons customers in pilot today say they chose us.

Founded by researchers.
Deployed by practitioners.

Struxiva is a research-driven, high-tech company built at the intersection of building physics, applied AI, and large-scale data engineering — created by practitioners who have owned the problems we solve. Our approach is grounded in real-world research, from field studies and simulation models to large-scale audit data and operator insight. We don't build features and look for applications — we start from validated evidence and work backwards to the product. The result is technology that reflects how buildings actually perform, delivering reliable, scalable solutions grounded in data, not assumptions.

Scientific lead
PhD in building energy systems
Published research in fault detection, thermal modelling, and HVAC optimisation. Scientific methodology underlies every detection rule in the pipeline.
Engineering lead
Production AI across regulated industries
AI and data infrastructure shipped in healthcare, energy, and regulated SaaS. Architect of the durable workflow pipeline and BRICK mapping layer.
Infrastructure lead
Large-scale data systems & cloud
Designed the multi-tenant storage layer and the stateless, multi-instance deployment model. Built for the data volumes and regulatory requirements of Nordic public-sector portfolios.