Hello everyone,
The webinar “Developing a System Engineering Knowledge Partner” raised many questions.
We would like to warmly thank Tony Komar, from Siemens DISW, for taking the time to answer them in detail during his presentation: Capella Days 2025 | Learn from Capella ecosystem members
We would also like to thank all the participants for teh rich discussions and relevant questions addressed to the speakers. Below are the answers to the questions that could not be addressed live:
| Full name | Question | Answer |
|---|---|---|
| Terre R | Hello, question about siemens and capella: what are the all the types of projects that Siemens uses capella for? | I work in Presales organization. I support customers use SMW which is based on Capella. In the process I built the landing gear model showed. So myself an other build models for demos, and I use is on coding project. I have not worked directly with any other Siemens organizations only many external customers. Other presales group in Europe may have, but not aware of it. Oh OK, I will send you an email during this week or the next week with following question, if it’s okay. Thanks a lot. |
| Terre R | Is it okay to send confidential data to External APIs (such as OpenAi APi) or is that okay? | It not Siemens policy to do this, so I don’t. We have internal models that allow it. |
| Hans-Günter Thonemann | How do you involve Engieers at SME companies for the capabilities you showed? | It currently case be case basis. I offer basic Capella training and show some capabilities around leveraging models. Some are not ready for AI, some are! Varying readiness is also my observation. Most are still far away from MBSE AI agents, though some isolated personal awareness exists. |
| Hans-Günter Thonemann | Did you study any effects of your AI prompts and Agent Concepts with the different LLMs other than that of Open AI’s ChatGPT | Not extensively. Currently the code is wired to open ai api’s it would be pretty trivial to use other model, since api calls via python are becoming very common. The RAG solution I have allows you to switch models in the notebook calls so comparision would be easy. Thanks for the answer and for your entire presentation. I am still at the non-agent-level of prompt experiments with different LLMs for the creation of Capella model artefacts related to specific project needs. |
| Mohammadreza Sadri | Hello, do you have any suggestion for resources and referances for someone who is interested to follow up on this approach? | Look at my last slide for some great papers. I can share the RAG solution. The agent solution should be able to review in a month or two. |
| John Roy Ballossini Dommett | I understand you’re working within the framework of a large corporation like Siemens and have access to or possibly even built Simcenter Studio for the very purpose of LLM systems architecture. As an MBSE student, I’ve been tinkering with locally hosted N8N automation workflows and free LLMs APIs such as Groq because proprietary/legacy paid versions are out of my reach… aside from your book and youtube - What would you say is the best pathway for 1) educating ourselves on LLMs SysML and 2) creating or accessing existing plugins or repositories for Open Source LLM for MBSE? Closely following your work, so thank you for this presentation! | 1) educating ourselves on LLMs SysML and 2) creating or accessing existing plugins or repositories for Open Source LLM for MBSE? If it is open souce, use a LLM tool. I found Chat GPT very good at pushing boundaries and learning. |
| Terre R | What is the syetem prompt you use? What are the LLMs used? How big they are? | They are quite large. The Contract noted feed into a system prompt. I change the system prompt based on the state of LLM and plan to use different contracts for the different hats a system engineer wears. Very interesting, so you adapt it to the system. Thanks |
| Terre R | Rewritten: did Siemens have an issue with sending data to external APIs, and have you considered using local LLMs instead? | The solution is able to work with different LLMs. I have access to internal to Siemens and ones and external . Interesting, thank you for this answer. |