Gavro SkyTech — Asset Intelligence & Aerial Inspection
THE GAVRO INTELLIGENCE PLATFORM

Aerial data is everywhere.
Actionable intelligence isn’t.

Most aerial inspection ends with a folder of images and a PDF. Ours ends with graded findings, a measurable digital twin and a decision your engineers can defend. The difference is the platform between capture and delivery — built, hosted and governed by us, in Australia.

DESIGN PRINCIPLE
Deterministic where it matters. AI where it helps.

Vision models are not deterministic. Ask one to measure the same thermal anomaly twice and you can get two different answers — which is fine for a photo caption and unacceptable for an asset decision. So we split the work: physical detection is computed deterministically from real measurement data; the AI model is restricted to classification and description of what has already been detected. The measurement is never the model’s opinion.

What this means in practice
Run the same thermal dataset through the platform twice and you get the same regions, the same ΔT values and the same finding count. Repeatability is a precondition for defensibility — if a finding can’t be reproduced, it can’t be relied upon in a maintenance or insurance decision.
Why competitors can’t claim it
Passing imagery to a general-purpose AI and publishing whatever comes back is fast to build and impossible to defend. Deterministic detection requires the measurement data itself — radiometric grids, not JPEGs — and the engineering to process it.
THE PIPELINE
Seven stages. Every engagement. No exceptions.

The same governed pipeline runs whether the asset is a strata roof or a water treatment precinct.

01
Capture
Survey-grade acquisition to a documented field methodology — saved missions, RTK-corrected positioning, radiometric thermal captured as true temperature data rather than false-colour images.
RTK/NRTK ± 2–3 cm · GDA2020 / MGA · radiometric R-JPEG · full flight records retained
02
Analyse
Photogrammetric reconstruction and radiometric extraction convert raw capture into measurement-grade data products — orthomosaics, point clouds and per-pixel temperature grids.
Photogrammetry · point cloud generation · DJI Thermal SDK per-pixel °C extraction
03
Detect
Deterministic detection identifies physical anomalies from the measurement data. Thermal analysis follows IEC 62446-3 methodology using neighbouring-panel local references. AI assigns labels to pre-detected regions — it does not decide what exists.
numpy/scipy region detection · IEC 62446-3 local references · multi-pass consensus on classification
04
Prioritise
Findings are graded against a defined framework and ranked by risk and operational impact — so the register tells you what to do first, not just what exists.
Five-point condition & severity scales · duplicate detection · risk-weighted ranking
05
Report
Executive-ready reporting: summary, defect register, annotated evidence imagery, and maintenance direction — formatted for the people who allocate budget.
Branded PDF generation · annotated evidence · exportable registers
06
Digital Twin
The asset is published as an interactive, measurable 3D model behind a tokenised client link — with key assets labelled and critical dimensions baked in.
Point cloud → streaming octree · browser viewer · measurement & section tools · DXF export
07
Ongoing Intelligence
Repeat cycles re-fly identical missions, so each capture extends a comparable history. Change becomes visible; trends become actionable; the asset record compounds in value.
Saved repeatable missions · epoch comparison · twin updates per cycle
GOVERNANCE & DATA HANDLING
Your asset data, under Australian control

The platform is self-hosted on Australian infrastructure and operated by a CISA/CISM-certified practitioner. There is no third-party cloud in the delivery path, no offshore processing, and no vendor with rights over your data. For critical infrastructure operators and SOCI-regulated entities, this is not a feature — it is a precondition.

Self-hosted, single jurisdiction
Capture, processing, storage and delivery all occur on infrastructure we own and operate in Australia. No data leaves the country as part of our workflow.
Access control & audit
Every operator action is logged — who ran what, when, for which client. Client links are tokenised, scoped to a single asset, and revocable on request.
Governed by certified practice
Data handling, chain of custody and confidentiality follow CISA and CISM-certified information security practice — held since 2011, not a marketing claim.
Presentable in your governance
Our architecture, retention posture and access model are documented and can be provided for your due diligence or SOCI data governance schedules.
OPEN OUTPUTS
Built to feed your systems, not lock you into ours

Intelligence that can’t reach your asset management system isn’t intelligence — it’s a PDF. Every deliverable is produced in open, standard formats aligned to the systems asset owners actually run.

Spatial
GDA2020 / MGA orthomosaics, DEMs and point clouds — GIS-ready on delivery
CAD
DXF export of twin measurements and sections, straight into your drawing environment
Structured data
Finding registers with grades, priorities and locations — exportable into EAM workflows
Roadmap
Direct GIS (Esri) and asset-management (EAM) integration on the platform roadmap
WHY IT MATTERS
The aircraft are the acquisition layer.
The intelligence is the product.

Any operator can buy the same aircraft. What can’t be bought off a shelf is a governed pipeline that turns capture into defensible decisions — and an operating model that produces the same rigour on the hundredth engagement as on the first.