AI Solution Architect

At a glance

Target group

Employees from (preferably) technical areas with an engineering background (degree).

Start

October 2026

Duration

1 semester

Scope

  • 5 ECTS (= 125 hours, of which 31 hours attendance, 94 hours self-study incl. project)
  • 10 live online learning units of 90 minutes each (= 15 hours)
  • 2 day workshops in presence á 8 hours (= 16 hours)
  • 0.5 days Final examination Project presentations (20 min. presentation per participant)

Conclusion

Further training certificate in accordance with Art. 78 Para. 1 No. 2b BayHIG

Fees

2.935,00€ p.p.

Speakers

Prof. Dr. Thomas Wieland (HS Coburg), Thomas Knorr (datadice GmbH)

Special feature

The course is run in cooperation with Zukunft Coburg Digital.

Course content

The AI ecosystem & strategic selection

  • Content: Current LLM landscape (Gemini, GPT, Claude, Open Source). Selection criteria: Latency vs. quality vs. cost.
  • Practical task: Creation of a decision matrix for a specific company scenario (When do I use a small local model, when a powerful cloud LLM?).

Architecture of language models & system prompts

  • Content: Token management, Context Windows and the art of "Programmatic Prompting". Building robust system instructions (guardrails) that do not "break out".
  • Practical task: Design a system prompt for a company assistant that strictly adheres to internal guidelines and blocks unwanted topics.

Local AI & data protection infrastructure

  • Content: Deep dive into local solutions (Ollama, LM Studio). Hardware check: What does a company laptop/server need?
  • Practical task: Installation and local test of a model; measurement of the response speed with different hardware settings.

Enterprise Integration & Copilots

  • Content: Technical control of Microsoft 365 Copilot & Google Workspace. API connections (custom GPTs / actions).
  • Practical task: Configuration of an "action" (interface) with which a chatbot can retrieve data live from an Excel cloud table.

IT security, ethics & rollout management

  • Content: Security (prompt injection), EU AI Act (technical documentation obligations). Change management: How do I involve the IT department and users?
  • Practical task: Creation of a technical compliance checklist for the rollout of a new AI tool in the company.

Knowledge management with RAG (Part 1: Data pipeline)

  • Content: How does company data get into the AI's brain? Chunking (cutting up documents) and embeddings (vectors). Basics of vector databases.
  • Practice task: Preparing a document set: Strategic chunking (by paragraph vs. by page) and discuss the pros/cons for search quality.

Knowledge management with RAG (Part 2: Quality & Evaluation)

  • Content: Improving retrieval quality. How do I technically prevent hallucinations? Evaluation: How can I recognize that my system is responding "well"?
  • Practical task: Test run of a RAG pipeline: Comparing the AI answers with the original sources and identifying sources of error.

Agentic AI – from response to action

  • Content: Introduction to autonomous agents. "Function Calling" explained simply: How AI learns to use tools (e.g. CRM query, SAP entry).
  • Practical task: Definition of a simple "toolbox" for an AI (Which interfaces in the company are ready for AI automation?).

Low-code automation & n8n

  • Content: Introduction to n8n as the "glue" between AI and business software. Workflow design: From email to AI analysis to the finished process.
  • Practice task: Building an automated workflow that classifies a customer request and creates a personalized response draft.

The website of Coburg University of Applied Sciences was translated using translation software provided by a third-party provider such as DeepL. The official text is the German version of the website. No liability is assumed, either explicitly or implicitly, for the accuracy, reliability, or correctness of the translations into another language.