Why physics-grounded, interpretable models outperform black-box ML for operational decision-making — and how to compose them as dynamical subsystems inside an AI and simulation platform.
01
Modelling
AI · Simulation · Decisions
Where AI meets simulation.
Optimal Reality is the AI and simulation platform for building and deploying intelligent applications — from physics-grounded models to runtime-personalised interfaces.
Built with Deloitte · trusted across
- Defence
- Transport
- Energy
- Critical Infrastructure
- Robotics
The scope
Every connected system. One operating reality.
Optimal Reality is the substrate beneath the world's connected things — unifying the systems that live, move and work into a single, decision-ready model of reality.
web.core.optimalreality / Nexus
Scope
Live
Homes, buildings, cities, utilities, health, energy — instrumented and responsive.
Move
Transport, fleets, skies, space — coordinated across modes and operators.
Work
Factories, mining, agriculture, supply chain — optimised with physics-grounded models.
Sensors → models → simulations → operators — in one stack.
The Kits
DDK
Data Development Kit
Define schemas, ingest streams, and make real-world data first-class in your autonomy stack.
MDKModel Development Kit
Train, swap and deploy models — from perception to control — with full visibility into how they perform.
FDKFrontend Development Kit
Build operator-facing applications and dashboards on top of live autonomy data — without rewriting plumbing.
SCDKScenario Development Kit
Author the synthetic worlds your systems need to learn in — with reproducibility and edge-case coverage built in.
The Physical AI Loop
01
Connect & Digitize
Sensor feeds, telemetry and live signals — unified.
02
Store & Structure
Real-world data made queryable and useful.
03
Segment & Understand
Raw streams turned into reasoned signals.
04
Simulate & Train
Validate behaviour synthetically before the real world.
05
Deploy & Manage
Roll out, monitor and govern at scale.
06
Edge Inference
Run intelligence where the action is.
Early Release Notes
Subscribe
v0.7 Latest
Workflow fine-tuning & smarter task connections
This release focuses on giving you tighter control over autonomous workflow behaviour and reducing the friction in wiring tasks together.
- New Workflow parameter fine-tuning panel — dial in thresholds and tolerances directly from the FDK.
- Improved Smarter task input/output mapping; fewer manual connections required.
- Improved Run comparison view now supports side-by-side diffing of metric outputs.
- Fix Resolved an edge case where streaming subscriptions could drop on long-running sessions.
v0.6
Live data streaming & map customisation
Real-time data is now a first-class citizen, and the geospatial map view can be tailored to your operational environment.
- New Live data streaming — react to sensor and telemetry feeds as they arrive.
- New Map customisation: custom layers, terrain overlays and asset tracking.
- Improved AI cost tracking surfaces inference spend per project and per workflow.
v0.5
Custom function builder & state management
Build reusable logic without leaving the platform, and keep tighter tabs on every running process.
- New Custom function builder — author and reuse functions across projects.
- New Granular access controls with per-project role assignments.
- Improved State management surfaces the lifecycle of every autonomous process.
v0.4
Synthetic data templates & swappable models
- New Synthetic data templates for common autonomy scenarios.
- New Swappable AI models — plug in perception, planning or control modules without rewiring.
- Improved Smarter monitoring & alerts with automatic anomaly flagging.
Get Started
Ready to ship real-world autonomy?
We work with teams building the next generation of physical AI — from robotics startups to enterprise operations.