AI4MBSE

Ongoing SysDICE R&D

Project type · SysDICE R&D · client-confidential MedTech application

A Secure, Domain-Specific Assistant for Complex Models

An ongoing SysDICE R&D initiative developing an assistant for complex MBSE models that is secure, domain-specific and integrated with engineering workflows — targeted at regulated environments where generic AI tools fall short.

Project type

SysDICE R&D · client-confidential MedTech application

Industry

MedTech / regulated engineering

Engagement focus

Making complex system models easier to explore under regulated constraints

Status

Ongoing SysDICE R&D · client-confidential MedTech application

Project context

SysDICE is developing an assistant for complex MBSE models that is secure, domain-specific and integrated with engineering workflows — designed for regulated environments such as MedTech, where model complexity and compliance expectations rule out generic AI tools.

The challenge

Proprietary model formats and complex text representations, thousands of elements and varied relation types, slower AI adoption in regulated industries, and uneven modeling practices. A representative source model scale involves 15,000+ elements, 500+ relations, 124 stereotypes and 38 metaclasses.

What SysDICE delivered

Specialized integration beyond off-the-shelf LLM tools, tailored to complex MBSE models.
Model-aware AI architecture designed for proprietary model formats and rich relation structures.
Security and domain constraints appropriate for regulated engineering contexts.
Integration with MBSE workflows so the assistant meets engineers where they already work.

Value created

Easier exploration and understanding of large, complex system models.
More accessible engineering knowledge across modeling teams.
AI assistance designed around regulated-engineering constraints rather than retrofitted to them.

Capabilities applied

LLM integration

Model-aware retrieval

Regulated engineering

Security-aware AI

MBSE workflows

Systems-model navigation

Discuss a similar project

Bring your challenge — we’ll help you shape a pragmatic MBSE and AI4MBSE path forward.

Project journey

How the project unfolded

A step-by-step view of the path from framing the problem to establishing a way forward.

01

Step 1

Understand model complexity

Characterise proprietary formats, thousands of elements and rich relation structures in real MBSE models.

02

Step 2

Define constraints

Set out the security, domain and regulatory constraints the assistant must respect from day one.

03

Step 3

Build model-aware integration

Develop an architecture that goes beyond off-the-shelf LLM tools to reason about MBSE models directly.

04

Step 4

Connect workflows

Integrate the assistant with the MBSE workflows engineers already use, rather than sitting alongside them.

05

Step 5

Validate with complex models

Exercise the assistant against representative model scale and complexity to refine behaviour over time.

Consulting · Implementation · Training

How this project connects to SysDICE's way of working

Every SysDICE engagement blends consulting, implementation and training. Here’s how those three modes came together on this project.

Consulting

Framing where AI can responsibly assist engineering in regulated environments — and where it cannot yet.

Implementation

Ongoing development of a model-aware, security-aware AI assistant integrated with MBSE workflows.

Training

Emerging patterns for helping engineers work with AI assistance around complex, regulated models.

Have a complex engineering project to discuss?

Start a conversation — no pitch deck required. Tell us where you are and where you want to go.