Platform · the Aegis engine
From live data to ranked scenarios.
Aegis connects the feeds you already produce, models your territory or organization as a network, predicts how failures travel through it, and ranks what deserves attention first. Around the engine sits a modular, scalable platform built by risk professionals who have done this work by hand and know how much time it consumes.
It starts with what you already have.
CCTV, drones, weather, river levels, traffic, air quality, satellite, on-site sensors: Aegis plugs into the feeds a territory or a plant already runs and merges them into one picture. Hover a feed to see what it brings.
CCTV
Fixed camera network: eyes on streets, sites and perimeters.
Drones
Aerial and drone imagery: a view from above, on demand.
Weather
Forecasts and radar: wind, rain and what is coming.
River levels
Gauges and water levels: how high, and how fast.
Traffic
Road-network flow: congestion, closures and access.
Air quality
Monitors: what is in the air, and where.
Sensors
On-site instruments: pressure, temperature, flow and more.
Satellite
Earth observation: wide-area imagery and change over time.
There is no new sensor network to buy. Aegis reads what the existing feeds report, whether that is an odd sensor reading, a camera frame or a field message, and holds it against the risk model of the system. Every team works from the same picture, and the picture is current.
Field teams
positions, tasks and reports in the same picture as the risk
Operations rooms
the live state of every scenario during an event, updated as feeds arrive
Corporate & EHS
traceability, ISO frameworks and compliance served from the same model
External stakeholders
insurers, agencies and neighbours briefed from one source
A live build over London: every marker is a public traffic camera, and the pinned panels stream weather, air quality and river data into the same workspace.
Your system, and the paths risk can travel through it.
Aegis starts from what is actually there: assets, functions, dependencies, and the interaction network that connects them.
Process units, substations, hospitals, storage farms, rivers, roads, populations: every entity carries its role, its state, and its dependencies. When the system changes, the risk picture changes with it.
On top of that model, a transparent algorithm pairs each hazardous source with each entity it can reach, and classifies every link by type, space and time. The result is an interaction network of risk pairs: the complete set of questions the engine has to answer.
The territory inventoried by an area scan: every facility found, classified by criticality, and opened here on a hospital's asset details.
The interaction network
Every hazardous source, every entity it can reach, and the path that carries the hazard between them. In the city model beside this, those become typed links, each drawn in its own colour: a plant can set off a substation by mechanism, a hospital depends on that substation for power, and a residential district sits inside the plant's reach by vicinity. Coverage comes from the algorithm, not from what a workshop happened to remember.
Every source, everything it can reach, and how the hazard gets there. One colour per link type.
A Risk Encoder that reasons like an experienced engineer.
Every pair in the network is a question: what can this source do to this target? The Risk Encoder answers from mechanism, the way an engineer would, not from a lookup of past events. Trained on more than 500,000 complete accident analyses from 85+ countries, it also anticipates scenarios no register has recorded yet.
One event branches into every scenario it can set off. Branches above the threshold are kept and grow; low-probability ones wait in reserve.
How to read this pass
- 01
An initiating event enters the network. Here: a gas release on a process unit.
- 02
For each scenario and the data available, the engine picks the physics micromodel with the best fit, then scores every possible outcome.
- 03
Scenarios above the threshold strike their target, and the struck target becomes the next source. The loop runs until the tree is complete.
- 04
Scenarios below the threshold are not discarded. They wait in reserve and are checked again the moment conditions change.
From thousands of scenarios to a short list.
A complete pass produces more scenarios than any team can read. Aegis ranks them by assessed severity, shows which links carry them, and points at the places where a barrier breaks the chain.
One selected substation: everything it feeds, everything that feeds it, and how deep a failure travels in each direction.
This is what acting on the ranking looks like. Select any entity and Aegis traces its dependency depth in both directions; the node panel shows its degree, its risk score and its place in the ranking. Reinforcement budgets and attention go where the analysis points, and every number can be traced back to the assumptions behind it.
Scenario trees
full propagation chains per site or territory
Ranked exposure
assets and zones ordered by assessed severity
Intervention points
where a barrier breaks the chain
Audit trails
every scenario traced back to its links, models and assumptions
Test a plan, see what it saves.
Connected to a reinforcement-learning module, Aegis generates and stress-tests mitigation plans against the same scenario tree, within your real constraints: resources, time, organization. The goal is simple: the barriers that stop the most, for the least.
Aegis tries a barrier on each link and keeps the ones that save the most. Here, a single well-placed barrier turns a full outage into a contained one. · in development
Every candidate plan is scored against the same scenario tree as the analysis, so a plan that protects one asset cannot silently expose another.
The mitigation module is in development with our design partners. We show it here because it is where the engine is going; this section will be updated as the module matures.
