Platform

A smarter way to understand infrastructure

Phyll transforms imagery into intelligence. Continuous. Persistent. Actionable.

How it works

From raw footage to persistent infrastructure memory

Every frame becomes an observation. Every observation resolves to an entity. Every entity accumulates state. Every state change becomes a delta. The result: an auditable system of record for the physical world.

CaptureObservationEntityStateDeltaScene
01

Capture

Ingest heterogeneous imagery from any existing source — dashcams, body cameras, drones, CCTV networks, phones, or fleet vehicles. No new hardware, no disruption to existing operations.

DashcamsDronesCCTVPhonesFleetsSatellite / aerial
02

Understand

AI detection and condition models process footage by source type and angle. Different model paths for ground-level vs. aerial vs. oblique imagery. Proprietary datasets for infrastructure-specific asset classes and defect conditions not covered by generic models.

Asset detectionCondition classificationChange detectionDamage IDCustom defect models
03

Remember

Each detected asset is resolved against existing records using geolocation, appearance, geometry, OCR/symbols, and temporal priors. The same real-world asset is recognised across different cameras, angles, seasons, and time — and assigned a durable persistent ID. State-to-state deltas are tied to the asset itself, not just frame comparisons.

Persistent asset IDsCross-source identityTemporal memoryState deltasEvidence ledger
04

Act

Intelligence flows directly into the tools teams already use — GIS platforms, asset management systems, risk dashboards, underwriting workflows, maintenance queues, and operations consoles.

GIS integrationAsset managementRisk dashboardsAPI / GeoJSON / CSVWork orders

Coverage

The cameras are already there.

Contracted camera partners give Phyll on-demand reach across the US and EU — no deployment, no new hardware.

Request coverage →
100K+
vehicles in Phyll's contracted camera network
80%
of US & EU addresses passed by the network every month
72 hrs
from coverage request to delivered intelligence

Economics

How the math works

Infrastructure intelligence at a cost that makes continuous monitoring feasible at scale.

<$0.005
per asset observed
11,000+
data points extracted per hour of video

Cost per mile — relative comparison

Phyll (dashcam-based)<$4/mi
Mobile mapping services$45–70/mi
LiDAR survey vehicles$70+/mi
Traditional manual inspection$200+/mi

50–100× lower cost per mile than traditional inspection methods

Key outputs

Decision-ready intelligence

Risk Alerts

Real-time notification of new hazards, damage, or condition changes on monitored assets.

Asset Condition Monitoring

Continuous health scores and condition trends for every tracked asset.

Maintenance Prioritisation

Ranked queues of assets requiring attention, ordered by urgency and risk.

Infrastructure Intelligence Dashboards

Unified views of asset health, change history, and exposure across portfolios.

Change Detection Reports

Before/after comparisons for any asset, corridor, or site at any point in time.

Historical Infrastructure Records

A permanent, queryable record of every asset state and event over time.

Why it's hard to replicate

Detection is the commodity. Memory is the moat.

Any team can run a detector. No team can instantly copy historical footage, repeat-pass coverage, customer workflows, asset-level histories, and years of temporal signal.

Proprietary footage access

Fleets and operators capture the world but overwrite it every 30 days. Once Phyll stores it, that historical moment is proprietary inventory that can't be retroactively sourced.

Evidence-linked observations

Raw footage becomes structured observations with source, time, location, confidence, model version, and provenance — not just detections.

Persistent asset identity

The same real-world asset is recognised across sensor type, angle, time, condition, and capture quality using a multi-signal match graph.

Asset-centric delta ledger

The valuable object is the asset history: state-to-state changes tied to asset_id with evidence and auditability — not isolated frame comparisons.

Compounding data flywheel

More footage → better models → better intelligence → more customers → more footage. Every observation makes the system harder to replicate.

Prediction layer

Temporal histories become training data for deterioration forecasting, risk progression, outage likelihood, and maintenance prioritisation.

Data & Security

We are not building Big Brother.

Phyll delivers infrastructure asset intelligence — not surveillance. We process footage for physical asset conditions and defects. People, vehicles, and non-infrastructure entities are not our product.

PII is handled at ingest, not post-process. We don't retain footage beyond what's required to generate asset intelligence. Customer data never leaves permissioned environments.

Traditional systems

Capture & retain everything

  • Faces & identities
  • License plates
  • Pedestrians
  • Vehicles & movement
  • Everything in frame

PII retained in the record

Phyll

Infrastructure assets only

  • Hydrants & utility poles
  • Signs & signals
  • Road & pavement condition
  • Roofs & structures
  • Guardrails & barriers

PII discarded at ingest

Data ownership guaranteed

Your footage and derived intelligence remain yours. Phyll holds no ownership claims on customer inputs or outputs.

PII handled at ingest

Personal data (faces, plates) is processed at the point of ingest — not retained as part of the asset record.

Permissioned access controls

Role-based access, audit logs, and permissioned data sharing across all customer environments.

Private deployment available

Phyll can operate within your infrastructure — on-premise or in your cloud tenancy — with no external data movement.

Infrastructure-only outputs

Deliverables are structured data about physical assets: conditions, locations, change deltas. Not footage. Not surveillance.

Compliance roadmap

SOC 2, ISO 27001, and sector-specific compliance (NIST, insurance data standards) on the product roadmap.

Source frame · dashcam pass

Dashcam frame from a pilot corridor with a detected stop sign
stop_sign 0.94

Structured output

{
  "asset_id": "3a5d5884",
  "class": "stop_sign",
  "condition": "Good",
  "location": { "lat": 32.8452,
                "lon": -117.0151 },
  "confidence": 0.94,
  "source": "dashcam",
  "captured": "2026-05-15",
  "observations": 3
}

Ready to run a pilot?

Upload footage from a corridor, site, or fleet. Receive map outputs, detections, and change reports within days.

Talk to us about a pilot