Four modules
that pay for
each other.

Most travel tools bolt detection, search and booking together and hope. ViaMe runs them as one feedback loop: the prediction model weights the routing graph, routing constrains what booking can attempt, booking outcomes retrain prediction, and the device throttles its own sensors to the live risk score.

ARINC 834 GTFS-RT NWP radar Social NLP On-device ARM inference
M·01 PREDICT M·02 ROUTE M·03 BOOK M·04 Disruption ML risk score r ∈ [0,1] Multimodal A* n-best diverse Saga orchestration all-or-nothing Edge adaptive outcomes retrain prediction device throttles sensors to live risk

It sees the cell
before the gate does.

A transformer-class model fuses four asynchronous feeds into a single disruption probability per itinerary leg, refreshed continuously. Because the feeds disagree in useful ways — radar leads the airline's own status board by minutes — the fused score crosses the alert threshold while there is still inventory to rebook into.

Validated against historical disruption sets at a minimum 85% accuracy on the binary "will this leg break" task.

Feed 01 · aviation
ARINC 834
Gate push-back deltas, equipment swaps and crew-timeout signals across trailing flights.
Feed 02 · ground
GTFS-RT
Real-time rail and transit positions, headways and platform changes for the surface legs.
Feed 03 · weather
NWP radar
Numerical weather prediction and live radar cells mapped onto the corridor and approach.
Feed 04 · crowd
Social NLP
Natural-language clustering of public posts; a "delay" cluster crossing 3σ is an early tell.
ArchitectureTransformer-class, ~10⁸ parametersmulti-head attention over time-aligned multimodal tokens
OutputPer-leg disruption probability r ∈ [0,1]alert threshold τ = 0.75, tunable per traveller profile
RefreshContinuous · up to 50 Hz on active legs
Accuracy≥ 85% on held-out historical disruption sets

Every edge knows
its own risk.

Transport is modelled as a directed multigraph — airports, stations and kerbsides are nodes; flights, trains and ride-shares are parallel edges between them. Each edge carries time, cost and the live risk score from the prediction model, so a risk-weighted A* search naturally avoids the legs about to break.

Instead of one "best" path it returns an n-best diverse set, so the alternatives don't all fail together when the same storm closes them.

A B C D r .08 r .11 r .74 r .69
GraphDirected multigraphparallel modal edges between transfer nodes
SearchRisk-weighted A*transfer-aware admissible heuristic
Outputn-best diverse routesdecorrelated failure modes
Latency< 500 ms5+ leg international itineraries

The model rides
in your pocket.

A distilled copy of the prediction model runs on the phone, so the risk score keeps updating when connectivity drops over an ocean or in a tunnel. Getting a 10⁸-parameter teacher down to an ARM-friendly student is a three-step compression pipeline — and the same risk score then governs how often the device is allowed to wake its sensors.

01

Distil

A compact student network learns to mimic the full teacher's risk outputs, preserving the decision boundary that matters around τ.

02

Prune

Structured pruning removes attention heads and channels that don't move the score, shrinking the graph for mobile memory budgets.

03

Quantise

Weights drop to 8-bit integers for ARM NPU execution — sub-200 ms inference with negligible accuracy loss.

<200ms
on-device inference · ARM
−72%
power vs. constant polling
15s–4h
risk-gated GPS interval

The full architecture,
in one document.

The technical whitepaper covers the fusion model, the routing heuristic, the Saga compensation logic and the on-device governor in depth.