Engineering RAG-LLM-An AI-Enabled Model-Based Engineering Platform for Intelligent Product Development
Client-Facing Research White Paper
Integrating Sparx Enterprise Architect, RAG-LLM, MATLAB/Simulink, SystemC TLM, DevOps and Industrial IoT
Prepared by:
IAS-Research.com — Engineering Research & Technology
KeenComputer.com — Digital Transformation, AI & IT Infrastructure
KeenDirect.com — Digital Commerce & Technology Services
August 2026
Executive Summary
Engineering organizations are entering a new era in which artificial intelligence, model-based engineering, simulation, embedded computing, Industrial IoT and DevOps can be integrated into a single digital engineering lifecycle.
Most organizations, however, still operate these capabilities as separate toolchains.
Requirements may reside in an MBSE repository. Engineering models may be developed in MATLAB/Simulink. Embedded software may be maintained in Git. Hardware/software architectures may be modeled using SystemC. Test results may exist in CI systems, while operational information remains in databases, service records and field equipment.
The result is a fragmented engineering information environment.
This white paper proposes an Engineering RAG-LLM platform that creates an intelligent layer across these existing engineering disciplines.
At the center of the proposed architecture is Sparx Enterprise Architect (Sparx EA) as the MBSE and systems-architecture backbone. RAG provides controlled access to engineering knowledge. Large Language Models provide natural-language reasoning and orchestration. MATLAB and Simulink provide engineering computation and simulation. SystemC TLM provides virtual hardware/software prototyping. DevOps provides automated verification and deployment. Industrial IoT provides real-world operational feedback.
The resulting architecture establishes a continuous engineering digital thread:
Requirements → Architecture → Modeling → Simulation → Virtual Prototyping → Verification → Deployment → Field Data → Engineering Intelligence
The objective is not to replace engineers or established engineering tools.
Instead, the platform is designed to make engineering organizations:
- faster;
- more knowledge-driven;
- more traceable;
- more automated;
- more consistent; and
- better able to reuse engineering expertise.
1. The Business Challenge
Modern engineering products are increasingly software-defined, connected and computationally complex.
Examples include:
- electric and hybrid vehicles;
- automotive ECUs;
- smart inverters;
- distributed energy resources;
- industrial automation systems;
- connected machinery;
- battery systems;
- embedded AI devices;
- Industrial IoT platforms.
Developing these products requires multiple engineering disciplines.
A typical product-development lifecycle may involve:
Systems Engineering → Electrical Engineering → Controls → Embedded Software → Hardware → Simulation → Testing → DevOps → Operations
Each discipline generates valuable information.
The challenge is connecting that information.
The engineering information problem
A typical engineering organization may have:
- thousands of requirements;
- hundreds of design documents;
- MATLAB scripts;
- Simulink models;
- source-code repositories;
- test reports;
- SystemC models;
- specifications;
- standards;
- service manuals;
- field data;
- engineering emails and reports.
Finding the right information can consume significant engineering time.
More importantly, information may be available but not connected to the engineering context in which it is needed.
2. The Proposed Solution
The proposed Engineering RAG-LLM platform creates an AI-enabled engineering layer around the existing engineering toolchain.
Core architecture
ENGINEER | v +----------------------+ | Engineering AI Portal| +----------+-----------+ | v +----------------+ | RAG-LLM | | Agent Platform | +-------+--------+ | +--------------+--------------+ | | | v v v Engineering Sparx EA Operational Knowledge MBSE Data | | | +--------------+--------------+ | v Engineering Context | +--------------+--------------+ | | | v v v MATLAB Simulink SystemC TLM | | | +--------------+--------------+ | v Verification | v DevOps | v Edge / IoT | v Field Data | +--------> RAG
This architecture transforms RAG-LLM from a general-purpose chatbot into an engineering intelligence platform.
3. Why Sparx Enterprise Architect Is Central
A key enhancement of the proposed architecture is the explicit role of Sparx Enterprise Architect.
Sparx EA serves as the Model-Based Systems Engineering backbone.
It can provide the structured engineering context required by the AI system, including:
- requirements;
- system architecture;
- subsystem architecture;
- interfaces;
- behaviors;
- relationships;
- traceability;
- verification information;
- engineering baselines.
This is fundamentally different from simply indexing engineering PDFs.
The RAG system can potentially retrieve both:
unstructured engineering knowledge
and
structured MBSE information.
This enables much more meaningful engineering questions.
For example:
"Which requirements are allocated to the inverter controller?" "Which Simulink subsystem implements this requirement?" "Which verification tests are associated with this requirement?" "What components could be affected if this interface changes?"
The AI therefore operates within the system architecture, rather than outside it.
4. The Engineering Digital Thread
The ultimate objective is to establish a digital thread connecting the engineering lifecycle.
Stakeholder Need ↓ System Requirement ↓ Sparx EA / SysML ↓ System Architecture ↓ Functional Architecture ↓ Hardware / Software Allocation ↓ MATLAB / Simulink ↓ SystemC TLM ↓ Software / Firmware ↓ Verification ↓ DevOps / CI ↓ Deployment ↓ Field Data ↓ Engineering Knowledge ↓ RAG-LLM ↓ Engineering Decision
This digital thread is one of the principal differentiators of the proposed platform.
5. Engineering RAG-LLM Architecture
Traditional RAG is generally represented as:
Question ↓ Retrieve Documents ↓ LLM ↓ Answer
Engineering RAG expands this model.
Engineering Question ↓ Context Identification ↓ RAG Retrieval ↓ MBSE Retrieval ↓ Engineering Data ↓ AI Agent ↓ MATLAB / Simulink / SystemC ↓ Verification ↓ Evidence-Based Engineering Response
The LLM becomes an orchestration and reasoning layer.
It does not become the authority for numerical or safety-critical engineering decisions.
6. Separation of Responsibilities
A professional engineering AI architecture should clearly define responsibilities.
|
Technology |
Primary Responsibility |
|---|---|
|
Sparx EA |
Requirements, architecture and traceability |
|
RAG |
Retrieval of trusted engineering knowledge |
|
LLM |
Reasoning, explanation and orchestration |
|
MATLAB |
Numerical computation and engineering analysis |
|
Simulink |
System and control simulation |
|
SystemC TLM |
Virtual hardware/software prototyping |
|
ML |
Prediction and pattern recognition |
|
IoT |
Field-data acquisition |
|
DevOps |
Automated testing and deployment |
|
Engineer |
Judgment, approval and accountability |
This separation is critical for client adoption.
7. Model-Based Systems Engineering
MBSE provides the foundation for managing engineering complexity.
A recommended Sparx EA model can include:
Requirements
- stakeholder requirements;
- system requirements;
- subsystem requirements;
- performance requirements;
- safety requirements;
- cybersecurity requirements.
Architecture
- system blocks;
- subsystem blocks;
- interfaces;
- hardware;
- software;
- communication networks.
Behavior
- use cases;
- activities;
- state machines;
- sequences.
Verification
- test cases;
- verification methods;
- acceptance criteria;
- traceability.
The AI layer can use this structured information as engineering context.
8. AI-Assisted Requirements Engineering
Engineering RAG-LLM can assist engineers in analyzing requirements stored within the MBSE environment.
Potential capabilities include:
- requirement search;
- requirement classification;
- duplicate detection;
- ambiguity identification;
- traceability analysis;
- missing verification identification;
- impact analysis.
For example:
Engineer: "Find requirements without associated verification tests."
The system could retrieve the relevant MBSE relationships and produce a traceability-gap report.
The engineer remains responsible for approving any change.
9. AI-Assisted Architecture Analysis
Natural-language interaction can provide a new interface to the MBSE repository.
Examples:
"Show me the architecture of the vehicle diagnostic gateway." "Which components communicate through CAN?" "Which subsystem owns this requirement?" "What interfaces are affected by changing the communication protocol?"
The advantage is that the AI can retrieve answers from the client's actual engineering model rather than relying on generic pretrained knowledge.
10. MATLAB and Simulink Integration
MATLAB and Simulink provide the computational foundation.
MATLAB
Potential uses include:
- numerical analysis;
- optimization;
- signal processing;
- statistics;
- machine learning;
- feature extraction;
- engineering visualization.
Simulink
Potential uses include:
- control-system modeling;
- dynamic simulation;
- state machines;
- signal processing;
- algorithm development;
- testing;
- code generation.
The RAG-LLM layer can help engineers move between natural-language engineering questions and computational workflows.
11. Engineering AI + MATLAB
Consider a client question:
"Analyze the transient response of the inverter during a grid-voltage disturbance."
The AI workflow can be:
Question ↓ Retrieve Requirement ↓ Retrieve Inverter Architecture ↓ Retrieve Relevant Standards ↓ Identify Simulink Model ↓ Configure Simulation ↓ Run MATLAB/Simulink ↓ Analyze Results ↓ Generate Engineering Report
The LLM interprets and orchestrates.
MATLAB/Simulink performs the engineering computation.
12. SystemC TLM-2.0 Integration
SystemC TLM introduces a virtual hardware/software environment.
It can be used to model:
- processors;
- memory;
- buses;
- CAN controllers;
- communication interfaces;
- peripherals;
- hardware accelerators.
This allows engineers to validate software and architecture before physical hardware is available.
The architecture therefore becomes:
Sparx EA Architecture ↓ Hardware / Software Allocation ↓ SystemC TLM ↓ Virtual Platform ↓ Firmware ↓ Virtual Validation
This is particularly relevant to embedded systems and automotive development.
13. DevOps and Continuous Engineering
DevOps provides automation across the lifecycle.
A reference pipeline is:
Engineering Change ↓ Git ↓ CI Pipeline ↓ MBSE Checks ↓ MATLAB Tests ↓ Simulink Tests ↓ SystemC Build ↓ Unit Tests ↓ Integration Tests ↓ Container Build ↓ Edge Deployment
The objective is to turn engineering verification into a repeatable and continuously executed process.
14. Automotive Application
AI-Assisted CAN/OBD-II Diagnostics
Automotive diagnostics is an excellent candidate for the Engineering RAG-LLM architecture.
The platform can combine:
- CAN data;
- OBD-II data;
- DTC codes;
- freeze-frame information;
- service manuals;
- vehicle specifications;
- historical diagnostic cases;
- engineering models.
Reference architecture
Vehicle ↓ CAN / OBD-II ↓ Edge Gateway ↓ Data Processing ↓ MATLAB Analytics ↓ RAG Knowledge ↓ LLM Diagnostic Agent ↓ Technician
A technician might ask:
"What are the most likely causes of this DTC given the current sensor values?"
The platform can retrieve the relevant technical information and analyze available data.
The recommendation remains subject to professional diagnostic validation.
15. Automotive MBSE Integration
Sparx EA can provide the system context.
Vehicle Requirement ↓ Diagnostic Architecture ↓ ECU ↓ CAN Interface ↓ Diagnostic Software ↓ Simulink Model ↓ SystemC Virtual ECU ↓ Test Case
This creates traceability from requirement to diagnostic implementation and verification.
16. Smart Inverter and DER Application
A second major application is intelligent control and optimization of:
- solar PV;
- battery storage;
- smart inverters;
- microgrids;
- distributed energy resources.
Architecture
PV / Battery / Grid ↓ Smart Inverter ↓ Telemetry ↓ Edge Gateway ↓ RAG-LLM ↓ MATLAB / Simulink ↓ Engineering Recommendation ↓ Operator / Engineer
Potential knowledge sources include:
- grid codes;
- inverter specifications;
- operating procedures;
- historical telemetry;
- weather information;
- engineering research.
17. Grid-Forming Inverter Research
The same platform can support advanced research into:
- grid-forming control;
- droop control;
- virtual synchronous machines;
- voltage regulation;
- frequency support;
- weak-grid behavior;
- reactive-power control;
- islanded operation.
A research engineer could ask:
"Compare the transient performance of two grid-forming control strategies under weak-grid conditions."
The system could:
- retrieve relevant research;
- identify applicable models;
- configure simulations;
- execute MATLAB/Simulink experiments;
- compare performance;
- summarize results;
- preserve the experiment and evidence.
18. Industrial IoT
Engineering RAG-LLM can extend beyond development into operational systems.
Sensors ↓ Edge Device ↓ CAN / MQTT / OPC UA ↓ IoT Gateway ↓ Time-Series Database ↓ Analytics ↓ RAG-LLM ↓ Engineer / Operator
Applications include:
- motors;
- pumps;
- compressors;
- industrial drives;
- power converters;
- battery systems;
- manufacturing equipment.
19. Predictive Maintenance
The platform can combine machine-learning analysis with engineering knowledge.
Sensor Data ↓ Signal Processing ↓ Feature Extraction ↓ ML / Anomaly Detection ↓ RAG Retrieval ↓ LLM Interpretation ↓ Maintenance Recommendation
This can combine quantitative evidence with:
- equipment manuals;
- maintenance procedures;
- historical failures;
- engineering specifications.
20. Digital Twin
The architecture can evolve toward a digital-twin platform.
PHYSICAL SYSTEM | Sensors | v Digital Twin | +----------+----------+ | | v v MATLAB Simulink | | +----------+----------+ | v RAG-LLM | v Engineer
Potential applications include:
- predictive maintenance;
- performance optimization;
- fault diagnosis;
- virtual commissioning;
- scenario analysis.
21. Engineering Knowledge Management
A major client benefit is preservation of institutional knowledge.
Engineering organizations often possess decades of knowledge distributed among:
- experienced engineers;
- technical reports;
- project documentation;
- simulation files;
- service records;
- source code;
- test results.
RAG provides a mechanism for making this knowledge searchable and usable.
The objective is:
Turn organizational engineering knowledge into an accessible engineering asset.
22. Knowledge Governance
Not every document should have equal authority.
A recommended hierarchy is:
|
Priority |
Knowledge Source |
|---|---|
|
1 |
Applicable regulations and standards |
|
2 |
Approved specifications |
|
3 |
Manufacturer documentation |
|
4 |
Validated internal engineering knowledge |
|
5 |
Verified research |
|
6 |
Experimental data |
|
7 |
General reference material |
The RAG system should preserve:
- source;
- revision;
- date;
- project;
- owner;
- authority;
- status.
23. Security and Intellectual Property
Engineering organizations may consider their models, source code and design information highly confidential.
A client deployment should therefore address:
- identity management;
- access control;
- data segregation;
- audit logging;
- encryption;
- model governance;
- document permissions;
- API security.
Possible deployment models include:
Private Workstation
For individual research and development.
Private Server
For an engineering group.
Private Cloud
For enterprise deployment.
Hybrid Architecture
Sensitive engineering data remains under client control while approved external AI services are selectively used.
24. Human-in-the-Loop Engineering
The platform should follow a clear principle:
AI proposes. Engineering tools calculate. Simulation validates. Engineers approve.
AI can assist with:
- research;
- retrieval;
- analysis planning;
- coding;
- documentation;
- interpretation.
Engineering tools should perform:
- numerical calculations;
- simulation;
- formal testing;
- regression testing.
Engineers remain responsible for:
- engineering judgment;
- safety;
- design approval;
- operational decisions;
- regulatory compliance.
This separation is essential for responsible enterprise deployment.
25. Client Value Proposition
The platform is designed to deliver value in five dimensions.
Productivity
Reduce time spent:
- searching documentation;
- preparing simulations;
- analyzing results;
- producing reports.
Engineering Quality
Improve:
- traceability;
- repeatability;
- documentation;
- regression testing.
Knowledge Retention
Preserve institutional engineering knowledge.
Innovation
Accelerate:
- research;
- prototyping;
- simulation;
- virtual validation.
Operational Intelligence
Connect engineering models with real-world data.
26. Business Value Measurement
A client implementation should establish measurable KPIs.
Engineering Productivity
- engineering hours saved;
- search time reduction;
- simulation preparation time;
- documentation time.
Quality
- test coverage;
- traceability coverage;
- defect detection;
- regression success rate.
Operations
- diagnostic time;
- mean time to repair;
- anomaly detection;
- maintenance cost.
Business
- product-development cycle time;
- project cost;
- customer response time;
- engineering throughput.
27. Recommended Client Implementation Roadmap
A phased implementation reduces risk.
Phase 1 — Engineering Knowledge Assistant
Deploy:
RAG + LLM + Engineering Documents
Objective:
Create a secure engineering knowledge assistant.
Phase 2 — MBSE Integration
Add:
Sparx EA + SysML + Requirements Traceability
Objective:
Connect AI to the engineering architecture.
Phase 3 — MATLAB/Simulink
Add:
MATLAB + Simulink
Objective:
Enable AI-assisted engineering analysis and simulation.
Phase 4 — SystemC Virtual Platform
Add:
SystemC TLM + Virtual Hardware
Objective:
Enable hardware/software co-validation.
Phase 5 — DevOps
Add:
Git + CI/CD + Automated Testing
Objective:
Create continuous engineering verification.
Phase 6 — Industrial IoT
Add:
Edge + Sensors + Telemetry + RAG
Objective:
Connect engineering development with operational systems.
Phase 7 — Digital Twin
Integrate:
Physical System + Digital Model + AI + Field Data
Objective:
Create continuous engineering intelligence.
28. Recommended Initial Proof-of-Concepts
Rather than attempting an enterprise-wide transformation immediately, clients should select one measurable engineering problem.
Option A — Automotive
CAN/OBD-II Diagnostic RAG-LLM
Option B — Energy
Smart Inverter / DER Engineering Assistant
Option C — Power Electronics
Grid-Forming Inverter Research Assistant
Option D — Industrial
Predictive Maintenance Engineering Assistant
Option E — Embedded Systems
AI-Assisted Hardware/Software Virtual Prototyping
29. Role of IAS-Research.com
IAS-Research.com can serve as the engineering research and advanced-technology partner.
Potential areas include:
- MBSE;
- MATLAB/Simulink;
- SystemC/TLM;
- embedded systems;
- VLSI;
- power electronics;
- smart grids;
- grid-forming inverters;
- Industrial IoT;
- AI/ML;
- digital twins;
- hardware/software co-design.
The objective is to transform research concepts into engineering prototypes and validated proof-of-concepts.
30. Role of KeenComputer.com
KeenComputer.com can provide the technology infrastructure required to operationalize the platform.
Potential services include:
- Linux infrastructure;
- Docker;
- DevOps;
- AI infrastructure;
- RAG deployment;
- databases;
- cloud/private-cloud systems;
- web applications;
- cybersecurity;
- monitoring;
- SME digital transformation.
This provides the bridge between engineering R&D and production IT.
31. Role of KeenDirect.com
KeenDirect.com can extend engineering capabilities into customer-facing digital commerce and service platforms.
Potential applications include:
- B2B eCommerce;
- engineering product catalogs;
- spare parts;
- technical products;
- service portals;
- customer support;
- maintenance services;
- Magento-based commerce.
This creates a potential lifecycle:
Engineering → Product → Deployment → Service → Customer → Commerce
32. Integrated Business Model
IAS-RESEARCH.COM | Engineering R&D | v Engineering AI | +----------+----------+ | | v v KEENCOMPUTER.COM Engineering Clients | AI / IT / DevOps | v Production Systems | v KEENDIRECT.COM | B2B / eCommerce | v Customers / Users
This provides a complete pathway from engineering research to commercial deployment.
33. Competitive Differentiation
The proposed platform is differentiated by combining several disciplines that are frequently implemented separately.
Conventional AI Assistant
Documents → LLM → Answer
Engineering RAG
Documents → RAG → LLM → Engineering Answer
Proposed Engineering RAG-LLM
MBSE + RAG + LLM + MATLAB + Simulink + SystemC + DevOps + IoT
This provides a substantially richer engineering context.
34. Strategic Vision
The long-term vision is an AI-enabled engineering digital thread.
REQUIREMENTS ↓ SPARX EA ↓ ARCHITECTURE ↓ MATLAB / SIMULINK ↓ SYSTEMC TLM ↓ VERIFICATION ↓ DEVOPS ↓ EDGE / IoT ↓ FIELD DATA ↓ RAG ↓ LLM ↓ ENGINEERING INSIGHT ↓ NEW DESIGN ↓ SPARX EA
The lifecycle becomes continuous rather than linear.
35. Conclusion
Engineering organizations have an opportunity to move beyond isolated generative-AI experiments and develop a structured AI-enabled engineering lifecycle.
The proposed Engineering RAG-LLM platform combines:
Sparx Enterprise Architect for MBSE and engineering traceability;
RAG for trusted knowledge retrieval;
LLMs for natural-language reasoning and orchestration;
MATLAB/Simulink for engineering computation and simulation;
SystemC TLM for virtual hardware/software prototyping;
DevOps for continuous validation;
Industrial IoT for operational data;
and digital-twin technologies for continuous connection between physical and virtual systems.
The result is a platform designed to help engineering organizations:
- accelerate development;
- preserve engineering knowledge;
- improve traceability;
- reduce repetitive work;
- increase simulation and verification productivity;
- connect engineering development with field data;
- and create new AI-enabled engineering services.
The strategic proposition is simple:
Do not replace the engineering toolchain with AI. Connect the engineering toolchain with AI.
And the operating principle is:
Retrieve the knowledge. Model the system. Calculate with trusted tools. Simulate. Verify. Deploy. Learn from field data. Keep the engineer in control.
References
- Sparx Systems — Enterprise Architect and Model-Based Systems Engineering resources.
- MathWorks — MATLAB documentation and engineering-computing resources.
- MathWorks — Simulink documentation.
- MathWorks — Embedded Coder documentation.
- MathWorks — Simulink Test documentation.
- Accellera Systems Initiative — SystemC and Transaction-Level Modeling resources.
- Lewis, P. et al. — Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks, NeurIPS.
- Vaswani, A. et al. — Attention Is All You Need, NeurIPS.
- NIST — Artificial Intelligence Risk Management Framework.
- NIST — Cybersecurity Framework.
- Relevant ISO, IEC and IEEE standards applicable to the client's engineering domain.
- Client-specific engineering requirements, specifications, models, test results and validated technical documentation.
About the Solution Providers
IAS-Research.com
Engineering Research • Embedded Systems • Power Electronics • AI/ML • IoT • MBSE • Digital Engineering
KeenComputer.com
SME Digital Transformation • AI Infrastructure • Cloud • DevOps • Web • IT Solutions
KeenDirect.com
B2B eCommerce • Magento • Digital Product & Service Commerce
Client Engagement
The recommended next step is a focused Engineering RAG-LLM Discovery and Proof-of-Concept, beginning with one high-value engineering workflow and progressively integrating Sparx EA, RAG, MATLAB/Simulink, SystemC and DevOps as measurable business value is demonstrated.
From engineering knowledge to engineering intelligence.
From engineering intelligence to measurable business value.