中文

YAS platform

The operating system for machine risk.

YAS turns how a machine behaves into one clear, real-time risk score — so machines run safer and risk is priced better, with the licensed insurer keeping authority over pricing.

1.27M
trips scored
17M km
kilometres scored
441K
driving hours
8.9 TB
telematics archive

The agentic fleet

A fleet of agents behind every score.

Each step has its own agent: one captures every signal and puts it in context, one scores it, and one coaches safer behaviour and retunes the model. Signals go in; a standardized AURA score comes out — sharper with every trip.

Telemetry sources

  • Electric vehicles
  • Autonomous vehicles
  • Drones
  • Robots

  1. Telemetry Pipeline Agent

    Capture & context

    Powered by Events API & SDK + ODD Engine

    Captures every machine signal, then tags each one with the conditions it happened in.

  2. Risk Scoring Agent

    Score

    Powered by Scoring Model

    ~100 factors into one score — and shows what moved it.

  3. 0100

    AURA score

    Normalized signal

  4. Coaching / Calibration Agent

    Feedback loop

    Powered by coaching & real claims

    Coaches risky behaviour down and retunes the model against real claims — so the next score is sharper.

One score, many stakeholders

  • Faster operational action

    Risk behaviours surfaced and coached down — safer operation, lower loss.

    For Operators

  • Better-priced risk

    An explainable, factor-level risk view that informs pricing.

    For Underwriters & capital

  • Live portfolio risk

    Portfolio- and trip-level risk, monitored in near real time with alerts.

    For Risk teams

Events API & SDK

Every signal, captured the moment it happens.

A small device in the vehicle records how it's moving — motion and GPS, many times a second — and cleans up the data on the spot. In the cloud, each trip is matched to the live weather and the exact road it was driven on.

A city mapped with live telemetry, risk heatmaps and connected machines

Motion & orientation

In-vehicle IMU: forward, lateral and vertical acceleration, pitch / roll / yaw, rotation rates, G-force and heading.

Position & trip

GPS location, speed, course and altitude, with per-trip distance, duration and timezone.

Weather

Live rainfall, temperature, UV index, humidity and active weather warnings.

Road & route

Map-matched road class, surface, environment, speed limit, urban density, curvature and elevation.

Machine telemetry

Signals that stay close to the machine.

YAS captures four raw sources close to the source — motion, position, weather and road — and turns them into one behaviour-aware record.

Motion

Forward accelerationLateral accelerationVertical accelerationRotation & heading

Position

LocationSpeedCourseTrip shape

Weather

RainfallTemperatureUV indexHumidity

Road

Road classSurfaceSpeed limitEnvironment

Event SDK

From device to verified record, in a few lines.

The in-vehicle SDK captures motion and GPS at the edge, pre-processes and batches locally, then syncs securely — context, scoring and attestation happen downstream.

YasSDK · Swift
// Configure the in-vehicle edge SDK
let yas = Yas.configure(
  trackerId: "TRK-3194",
  vehicleId: "EV-0427"
)

// Start a trip — motion + GPS captured at the edge
let trip = yas.startTrip(type: .passenger)

// Samples stream as the machine moves
trip.onSample { s in
  // s.xyAcceleration · s.zAcceleration · s.location
}

// End → batched, synced, scored downstream
let record = await yas.endTrip(trip)
print(record.auraScore)   // 0–100

Data Context Annotation

Read risk in its true context.

Raw telemetry is meaningless without context. A hard brake is unremarkable on a dry highway and alarming on a wet downgrade in a school zone.

The ODD Engine tags every event with the exact conditions it happened in — road, speed, weather, density — so risk is read in context, not in the abstract. That history is also a moat: a newcomer can't go back and label road conditions it never recorded, and ours grows with every trip.

Operational Design Domain

  • Road class
  • Speed regime
  • Curvature
  • Gradient
  • Lanes
  • Surface
  • Urban context
  • Road environment
  • Traffic density
  • Weather
17M km
already ODD-annotated exposure

Risk factors we track

Speeding

  • Speeding frequency
  • Speeding intensity
  • Traffic-relative speeding
  • Environment-specific speeding
  • Speeding on curves

Longitudinal control

  • Braking frequency
  • Acceleration frequency
  • Grade-adjusted braking intensity
  • Slope-aware acceleration intensity
  • Rolling-stop frequency

Stability & manoeuvres

  • High G-force events
  • Sudden lane change
  • Driving stability (curvature)
  • Jerk smoothness

Driver state

  • Fatigue signals

Environmental & contextual risk

  • Road-surface risk
  • Operating-area density
  • Weather-related frequency
  • Sun-glare risk
  • High-risk-area geofences

Efficiency

  • Excessive idling
An electric delivery van, sidewalk robot and delivery drone sharing a city street at dusk
“A whole category of machines — autonomous fleets, robots, humanoids — was running without a real measure of its risk. YAS changes that: operators can finally see how safely each machine runs, cut losses, and earn the trust to scale.”
William LeeFounder, YAS

Scoring Model

Score any machine on one scale.

YAS breaks each trip into about a hundred risk factors and rolls them into one score — on the same scale for any vehicle, fleet, or machine, anywhere.

~100
normalized risk factors
17
safety parameters
807,480
EV fleet trips tracked

And it explains itself. When a score moves, YAS shows which behaviours moved it — not just that it changed. That "why" is what an underwriter can actually act on.

How the score is built

AURA score

1

Behaviour tracked

2

Decompose

3

Score each metric

4

Normalize score

5

Behaviour reward

6

Calibrate

Use cases

Five product views. One live signal.

Five product views show how YAS turns raw machine telemetry into risk intelligence — tracking machines, surfacing risk, and keeping fleet operators and insurer partners on the same signal.

01 / 05
Live Machine Tracking
Live machine tracking dashboard with active vehicles, route status, and fleet risk indicators.

Feature 01

Every active machine on one map

Delivery robots, EV taxis, AVs — with live speed, AURA score, and incident flags. Robot-07 stopped mid-route: velocity anomaly flagged, policy review triggered.

Cryptographic provenance

A verified record no one can dispute.

A score only counts if someone outside the operator will trust it. YAS turns every record into proof, with three guarantees.

Cryptographic provenance

A tamper-evident chain from capture to attestation. The record cannot be retroactively altered.

Structural neutrality

The operator cannot edit its own record. The firewall is architectural, not contractual.

Regulated standing

A regulated licence and carrier co-sign give the attestation a basis regulators already recognise.

An operator attesting to its own safety is marking its own homework. Neutrality is the point — and it is built into the architecture, not promised in a contract.

Governance

Trust is non-negotiable.

Regulators expect AI to be open, accountable and easy to explain. YAS was built that way from day one — not patched on later to pass an audit.

[Regulator-ready]

Built to meet what regulators expect of AI.

[Full audit trail]

Every input, score and explanation is time-stamped and kept.

[Plain-language reasons]

Each score comes with a clear explanation anyone can read.

[A human decides]

A qualified person always makes the final call.

[Fairness-tested]

Scores are checked so no group is treated unfairly.

Every score has a reason

Each score comes with a plain explanation of the exact behaviours behind it — operators, decision-makers and regulators can all follow it.

A person makes the call

YAS gives the risk picture; it never makes the decision. A qualified person reviews and signs off every call.

Changes are tested first

Before any model update goes live, it's tested against real operating data and must clear strict safety checks.

Nothing goes undocumented

Every input, score and recommendation is logged and time-stamped, so any decision can be traced and reviewed later.

In production

Trusted across the machine economy.

A JOIE electric taxi
JOIE EV TaxiFor EV fleet operators

Driver risk & safety monitoring

Each driver's AURA score surfaces risky behaviour and where to coach it down — safer fleets, lower loss.

Zurich headquarters
ZurichFor insurers

Dynamic, behaviour-based pricing

Cover is priced directly from the AURA score, so premiums track real behaviour rather than static proxies.

An autonomous service robot
AxonexFor robotics operators

Safety & loss reduction

A robot's AURA score shows how safely it operates — cutting incidents and loss, and giving new deployments a trusted track record from day one.

Get started

Put a verified risk signal behind every machine.

Talk to us about scoring your fleet, robots or autonomous systems — or see where YAS already runs.

Talk to us