DE·EN

Workshop

Java modernization with self-hosted AI.

Two weeks on your own codebase, separated by a break in which you ship what the first week produced. The AI models we use run on servers operated by vogella, or inside your own network. Your source code never reaches a public AI provider.

  • Length2 x 5 days, week 1 also bookable on its own
  • AudienceTeams with long-lived Java applications, tech leads, architects
  • ResultEverything lands in your repository, not in slides

Your code stays with you

For many organisations, AI adoption comes down to a single question: where does the source code go? In this workshop, nowhere.

  • Models run on servers operated by vogella in Germany, or entirely inside your own network or data centre
  • No code, no prompts, and no logs reach a public AI service
  • No training on your data, no retention beyond the workshop
  • On request we set the same stack up on your side, so your team can work independently between the blocks and afterwards

The two weeks

Topics agreed up front

There is no fixed day-by-day agenda. We agree up front which of the topics below actually pay off in your codebase, and work through those. What fits into a week depends on the state of the repository, and we tell you that before you book.

Week 1: build the foundation and ship it

  • Make the build reproducible and get CI green again
  • Characterization tests as a safety net before anything changes
  • Dead code, expired feature flags, unused dependencies
  • Move to a current Java version
  • Library updates in waves, green build after each wave
  • CVE scanning, security patches, SBOM and license review
  • Test coverage where the code actually changes
  • Flaky tests, CI runtime, quality gates
  • javax to jakarta, where that is due

By the end of the week everything is in a state you can ship, plus a prioritised backlog and a playbook with custom agent skills.

Week 2: build on the released state

  • Review: what the release showed, which regressions came up
  • Make module boundaries visible, break cycles
  • Find the seams for incremental extraction
  • ArchUnit rules so the structure does not erode again
  • Profiling with real production data instead of guessing
  • N+1 queries, memory use, virtual threads where they help
  • Logging, metrics, and tracing as lasting evidence
  • Incremental UI modernization following the strangler pattern
  • Handover: skills, conventions, and a prioritised roadmap

Where we start here is decided only after the release. That is why the week opens with a review rather than a prepared plan.

The break in between

4 to 6 weeks

The break is part of the programme, not a scheduling problem. A Java upgrade and thirty library updates are not verified in a workshop, they are verified in a real release: under load, with your data, with your QA. During that time your team takes the week 1 state through QA and production and works the backlog with the playbook on its own. That is the real test of whether the way of working holds up without us. Whatever surfaces shapes the agenda for week 2. We stay reachable for questions in between.

What your team has afterwards

In the repository

  • A green, reproducible build and a CI pipeline people trust again
  • A noticeably smaller codebase
  • A current Java version and current libraries, proven in production
  • Measurably higher test coverage from reviewed tests
  • Quality gates and AI code review on every pull request
  • Architecture rules that hold the cleaned-up structure in place

In the team

  • A migration playbook with custom agent skills
  • Evidence from the break that your team can work with it alone
  • A prioritised roadmap for what two blocks cannot cover
  • Remaining CVEs documented and assessed rather than unnoticed

Formats and prerequisites

Other formats

  • Week 1 on its own as an entry point, week 2 decided afterwards
  • A remote one-day assessment up front that shapes the agenda
  • Three days compact, focused on two topics
  • Half-day follow-up some weeks after block 2

Individual topics from above are also available as fixed-scope work with a fixed length and price, see AI Services.

Prerequisites

  • A Java codebase the group may work in, with access to the repository and CI
  • Solid Java experience on the team, no prior AI experience needed
  • A build that has run locally at least once
  • For the two-block format: a release window between the weeks and some team capacity for the backlog
  • Network access to the vogella AI server, or installation inside your own network, sorted out in advance

Onsite or online

Both blocks run at your site or fully remote, in English or German. We agree the topics together up front, after looking at the codebase. Dates and price on request.

Request this workshop

Or reach us directly: sales@vogella.com · +49 40 7880 4360