Engineering Intelligence · Platform Proposal
Engineering Intelligence Platform.
The AI-native operating layer for digital twins, robotics, and engineering systems.
Input Engineering Goal
→
The layer we build Engineering Intelligence
→
Output Engineering Decision
This proposal is much larger than an AI assistant. The project evolved from an "AI harness for simulation" concept into a full engineering platform. Frame the whole talk around this three-step flow — it returns on the closing slide.
TL;DR · the whole deck in 30 seconds
Not a harness. An Engineering Intelligence Platform .
01
Our IP is the Engineering Intelligence Kernel — not the simulator, not the LLM.
02
Built on NVIDIA Omniverse — we orchestrate physics, we don't reinvent it.
03
One kernel, every domain — robotics, manufacturing, energy, healthcare.
If they remember nothing else, these three points are the takeaway. The rest of the deck is the argument for them. Mentally return here whenever a technical detail threatens to lose the non-technical half of the room.
Act 1 · Mental model shift
We started with a narrower question.
Wrap a simulation engine with an AI harness so engineers can talk to it in natural language. Clean, useful — and, as we'll show, far too small.
Who User
→
Original concept AI Harness
→
Back end Simulation Engine
Be honest about the starting point. We originally viewed the solution as an AI harness sitting in front of a simulation engine. This architecture is too narrow: it optimizes a single tool instead of the engineering workflow around it.
The landscape
AI harnesses already exist — for everything except engineering.
Coding AI
Game Dev AI
Generic AI Platform
Simulation AI
Purpose
Write & ship software
Produce game assets
Build AI applications
Drive one simulator
Primary output
Code
Art / levels
Apps & agents
Simulation runs
Primary user
Developer
Game studio
Product team
Simulation engineer
Each harness creates an artifact. None of them orchestrate engineering.
Existing AI harnesses create artifacts — code, assets, apps, runs. They do not plan, reason about, or decide across the engineering lifecycle. That is the white space this proposal targets.
The gap
Engineering is missing its intelligence layer.
Top Engineering Goals
→
Absent today Missing Intelligence Layer
→
Tooling Simulation
Tooling Optimization
Tooling Visualization
Tooling Enterprise Data
→
Bottom Engineering Decisions
Engineering workflows require planning, simulation, and decision support — not just a chat box in front of a solver. Today the intelligence that ties goals to decisions is built by hand, every time, inside every team.
Act 1 · The shift
Not a harness. A platform.
Input Engineering Goals
↓
The platform Engineering Intelligence
↓
Simulation
Optimization
Visualization
Enterprise Data
↓
Output Engineering Decisions
This is the first "aha" moment. The intelligence layer sits between goals and decisions, orchestrating the four engineering disciplines below it. Hold here — let the audience absorb the shape before turning the page.
Act 2 · The platform
Three layers. One purpose.
Applications Domain experiences
Engineering Intelligence Platform Orchestration · reasoning · decisions
Simulation Runtime Solvers · physics · digital twins
Infrastructure GPU compute · storage · networking
Applications sit on intelligence. Intelligence sits on simulation. Simulation sits on infrastructure.
Walk the stack bottom-up. Infrastructure is generic. Simulation runtime is where existing engines live. The middle layer — the Engineering Intelligence Platform — is the one we own, and the subject of the next slide.
The heart of the platform
Engineering Intelligence Kernel.
Applications
↓
Our IP · The Kernel
Engineering Intelligence Kernel
Intent
Context
Knowledge
Planning
Workflow
Decision Intelligence
↓
Simulation Runtime · OpenUSD · Visualization · Enterprise Data
This is the most important slide in the deck. Every platform has a core: Linux has the Kernel, Kubernetes has the Control Plane, Omniverse has Kit. Our platform has the Engineering Intelligence Kernel. It is where our intellectual property resides — everything above and below can evolve; this stays.
Orchestration
From goal to engineering decision — not prompt to output.
01 Engineer
→
02 Goal
→
03 Kernel
→
04 Planner
05 Simulation
→
06 Optimization
→
07 Recommendation
→
08 Report
The kernel plans the workflow, dispatches the right solvers, and returns a recommendation the engineer can defend.
Contrast this with simple AI prompting. An LLM answers one question; the kernel orchestrates a whole workflow — planning what to simulate, in what order, against what constraints, then synthesizing a recommendation rather than a raw result.
Architecture · separation of concerns
Each layer is swappable. The kernel is not.
Applications Domain plug-ins
Engineering Intelligence Kernel LLMs · Knowledge Graph · Workflow Engine
Simulation Runtime Omniverse · Isaac · PhysX · RTX · CUDA
OpenUSD Universal scene description
Visualization BabylonJS · ThreeJS · Unity WebGL
Enterprise Data PLM · ERP · telemetry
Stress separation of concerns. We are not in the business of building new solvers, new renderers, or a new data lake. We integrate best-in-class runtimes and focus our engineering on the kernel.
Leverage · not reinvention
We don't compete with simulation engines. We orchestrate them.
Engineering Intelligence Kernel Ours
NVIDIA Omniverse Simulation platform
GPU Computing RTX · CUDA · DGX
Infrastructure Cloud · on-prem · hybrid
Omniverse gives us a physics-aware simulation runtime, OpenUSD interoperability, and a large ecosystem on day one — so we can spend our effort on intelligence, not plumbing.
The strategic argument: leverage versus reinvention. Building a competitive simulation stack would take a decade and miss the market. Building on Omniverse lets us ship intelligence while standing on NVIDIA's runtime, ecosystem, and roadmap.
Act 3 · Differentiation
A new category — by definition.
Coding AI
Game Dev AI
Generic AI Platform
Engineering Intelligence Platform
Primary user Developer Game studio Product team Engineer Core output Code Assets Apps Engineering decisions Engineering knowledge — — — Native Simulation — — — Orchestrated Optimization — — — Orchestrated Digital twins — — — First-class Decision support — — Bolt-on Core Enterprise data — — Bolt-on Core Workflow orchestration — — — Core
Make it obvious we are defining a new category, not entering an existing one. Every other AI platform is artifact-centric; only this one is decision-centric across the full engineering lifecycle. The highlighted column is the takeaway.
Where our intellectual property lives
The Kernel is the asset. Everything else can evolve.
Applications Swappable
Engineering Intelligence Kernel Our competitive advantage
Simulation Runtime Omniverse · swappable
OpenUSD Open standard
Visualization Open source
Infrastructure Commodity
If a better simulator appears tomorrow, we adopt it. The kernel remains ours.
Everything underneath the kernel can evolve — and will. The kernel remains our competitive advantage because it encodes engineering reasoning, not any particular runtime. This is the slide investors should photograph.
Extensibility
One kernel. Every engineering domain.
The Kernel
Engineering Intelligence
Extensibility through domain plug-ins. The kernel stays constant; each vertical supplies its own knowledge pack, solvers, and data connectors. One platform, many engineering domains — the long-term surface area.
Roadmap
From copilot to autonomous engineering.
Autonomous Engineering
Vision
Each step earns the next. The kernel is the turn that makes the platform — and autonomy — possible.
Gradual evolution. We start where the market is ready (a copilot), earn the right to extract the kernel, open it into a platform, and only then pursue autonomous engineering. We are not asking the audience to bet on autonomy today — we are asking them to bet on the path that leads there.
Closing
We are not building another AI harness.
We are building the Engineering Intelligence Platform.
Input Engineering Goal
→
Our IP Engineering Intelligence
→
Output Engineering Decision
End with confidence. Pause after the first line. The Engineering Intelligence Kernel becomes the operating layer connecting AI, simulation, optimization, visualization, and enterprise engineering systems. Leave the slide up; do not rush the Q&A.
1 / 16
← / → · scroll · swipe