Course
digicode: OSOW
Open Source & Open Weight: Models, Hosting, and AI Sovereignty
Course facts
Download as PDF- Understanding the differences between proprietary, open-source, and open-weight AI
- Overview of leading models (including Apertus, Mistral, DeepSeek, and Meta) and their benchmarks
- Recognizing «open-washing»: how to tell if an AI tool marketed as open is truly open
- Understanding cloud independence, data protection, and digital sovereignty in a corporate context
- Hands-on testing of open-source frameworks, on-premises setups, and specialized interfaces
- Evaluating costs, performance, and operational scenarios (on-premises vs. sovereign cloud)
- Compliance and regulatory framework (EU AI Act, Swiss data protection, and GDPR)
1 Fundamentals
- Definitions, Licenses, and Differences Between Commercial APIs and Open-Weight Models
- Why Open Source: Geopolitical Factors, Independence from the U.S. Cloud Act, and Cost Efficiency
- Model Overview and Benchmarks
- Focus on Switzerland: Apertus (ETH Zurich/EPFL) as an Example of a Fully Open, Sovereign AI «Made in Switzerland»
- Hands-On Exercise: Comparative Evaluation of Model Responses
2 Open-Source Ecosystems & Interfaces in Practice
- Applications and Frameworks (including Euria, Apertus, Mistral, Qwen, Meta, DeepSeek, Hugging Face)
- Case Study: When «Open» Isn’t Open: Using Lumo (Proton) as an Example to Critically Examine Routing Logic and System Prompts
- Local Execution vs. Sovereign Cloud Hosting (Switzerland/EU)
- Data-Protection-Compliant Integration into Existing Workflows
- Hands-On Exercise: Testing Various Open-Weight Models via Privacy-Friendly Interfaces
3 Prompting, RAG & MCP with Open Models
- Context Engineering and System Prompts for Open-Source LLMs
- Integrating your own company data via Retrieval-Augmented Generation (RAG) and Model Context Protocol (MCP)
- Fine-Tuning vs. RAG: When is it worth adjusting your own model weights?
- Hands-on exercise: Building a simple knowledge system based on an open-source model
4 Use Cases & Business Scenarios
- Processing Highly Sensitive Data (Legal, HR, Finance, Healthcare)
- Use of Open Source in the Public Sector and Regulated Industries
- Cost-Benefit Analysis: TCO (Total Cost of Ownership) of Proprietary vs. Self-Hosted AI
- Hands-On Exercise: Developing an Open-Source AI Use Case for Your Own Company
5 Governance, Law & Security
- Regulation & the EU AI Act: Exceptions and Requirements for Open-Source Models
- Data Protection and Professional Confidentiality: Secure Processing of Sensitive Data Without Third-Party Risk
- Security Aspects: Model Alignment, Guardrails, and Licensing Law for Open-Source Weights
- Responsible and Transparent Use in the Corporate Environment
6 Implementation, Strategy & Outlook
- Assessing Your Own AI Roadmap in Terms of Open-Source Potential
- Creating an Action Plan for Sovereign AI in Your Organization
- Future Trends: Open-Source Agents, Local On-Device Models, and European AI Sovereignty
- Introduction: Gathering experiences and questions
- Presentation: Teaching the basics, showing and explaining examples
- Interaction: Evaluating and discussing examples, developing your own ideas
- Practice: Functions, use cases, hands-on projects, infrastructure
This course is designed for decision-makers, IT managers, legal and compliance officers, and professionals working in data-sensitive fields. Among others, it is intended for:
- Leadership & IT Strategy: CTO, CIO, Chief of Staff, IT Manager
- Legal, Risk & Compliance: Head/Manager of Legal, Data Protection Officer (DPO)
- Public Sector & Healthcare: Digital Transformation Project Managers, Innovation Leads
- HR & Finance: Individuals Responsible for Confidential and Personal Data Processes
- Product & Engineering: AI Engineers, Software Architects, Product Owners
The following courses are also suitable for IT professionals who want to delve deeper into specific topics:
- Sovereign Cloud with Kubernetes
- Sovereign Platform Engineering with Kubernetes
- RAG Tech Deep Dive
- LLM Engineering & AIOps
Would you like to explore governance in greater depth? Then check out the following course:
There are no formal technical prerequisites for this course, and no programming knowledge is required. However, a basic understanding of AI applications—regardless of Open Source and Open Weight—is recommended.
Please bring your own laptop and make sure to set up the accounts we previously provided for the course day.
This will allow you to save what you see and learn directly in your own environment and immediately apply it to your daily work at the company.
If you do not have a laptop, we can provide you with a computer. Please contact info@digicomp.ch after registering for the course.