Local AI in Production — hosted by razzfazz.ai

This group is for people who run AI inside their own organisation: on their own hardware, in their own network, with no data leaving the building.

We meet four times a year, online, 90 minutes each. The focus is what sits between proof of concept and production: which models actually run on which hardware, how RAG behaves with real document collections, how to get authentication and permissions right, what GDPR and the EU AI Act require in practice, and where projects fail.

A recurring focus is software development and software testing. Both disciplines are where local models meet daily work, and both raise questions that generic AI talks skip: how to keep source code and test data inside the company, how to derive test cases from textual and graphical requirements and push them into Jira and Xray, how to support UAT without handing the business process to a cloud vendor, and how to give a model binding rules on compliance, security and code quality. We treat EitL (Expert-in-the-Loop) as the default design, not as a fallback.

Recurring topics:

• Software development: code assistance on local models, repository access, rules for compliance, security and code quality
• Software testing: test case derivation from requirements, Jira and Xray integration, test data, UAT support
• Model serving and hardware: what a single box delivers and where the limits are
• RAG in practice: document quality, chunking, retrieval
• Integration with existing systems via MCP and REST
• Governance: ISO/IEC 42001, EU AI Act, AI agreements with works councils and staff
• Operations: updates, air-gapped environments, support

Hosted by razzfazz.ai (SEQIS Group, Austria). We curated an open-source stack for local AI that is free for private use and evaluation. SEQIS has worked in software testing, QA and requirements engineering for 25 years, which is where the development and testing focus comes from. These sessions are not product demos. If you run a different stack or are still evaluating, you are welcome, and reports from the audience are what makes this group worth attending.

Free to attend. No recording, so people can speak openly.