AI in practice and platform engineering: trends at the OOP conference in Munich

As in the previous year, AI was one of the central topics at the OOP conference, which had its own dedicated track in the program. However, AI also played a central role in almost every one of the 150 other presentations and workshops. Artificial intelligence and GenAI have now reached the core of architecture and software projects, which was evident in all talks. According to organizer Sigs Datacom (a member of the heise group), around a thousand visitors came to Munich from 3 to 7 February to get an idea of this.

Tessa Pfattheicher and Lars Orta provided a vivid example of the problems encountered in AI practice in their talk about their experiences in developing a chat bot. The difficulties lay less in dealing with the model and the architectural design than in the unpredictability of the possible responses of the GenAI application and the resulting difficulties in understanding. This concerns both the systematic evaluability of these responses in an information technology context as well as the expectations of the customers and the communication of the team with the stakeholders.

Other presentations highlighted the connection between data and AI, quality assurance and the security of AI projects. Oliver Kolar, for example, shared tips on how AI can be used profitably in a company in compliance with data protection regulations. The current hype surrounding autonomous AI agents that make logical conclusions and autonomous decisions was not yet a major topic at the conference.

Platform engineering reduces the cognitive load

In addition to the classic development and architecture topics, platform engineering was a focal point of the program. Peter Diefenthäler, for example, vividly illustrated the many, widely ramified aspects that need to be considered when introducing platform engineering, including the various business areas, basic technologies and engineering cultures of the departments in larger companies. There are also considerations regarding tools, governance and costs. In his presentation, Markus Zimmermann added tips on practical procedures and recommended an internal evangelist to get developers excited about the new platform. All of this is intended to sort out the IT proliferation, bridge silos and reduce the cognitive burden on developers.

Microsoft and GitHub were given a keynote on the platform topic by the organizers. April Yoho and Julia Kordick presented the Dev Box and Azure Deployment Environment tools, which are now completely permeated by Copilot.

More legacy and less sustainability

Another omnipresent topic was how to deal with legacy: rebuild, redevelop or shut down? Michael Stal posed this question in his detailed overview presentation with many examples from Siemens practice. The problem with capturing legacy systems often lies in the fact that the old developers retire and leave behind poorly documented legacy systems. Stal recommends quickly tapping the remaining developers as a source of information. If this is no longer possible, AI helps to see through the legacy code.

Some presentations posed the question: Back to basics? – More REST? Design patterns? (Module) monolith? Sustainability in IT was less of a focus than in the previous year. Zorina Alliata and Hara Gavriliadi, for example, looked at how to develop and operate AI applications in an environmentally friendly way.

Many participants enjoyed the supporting program: Nightschools for late risers, a Pacha Kucha night, an unconference or an ultimate IT regulars' table. Two of the five conference days offered practical workshops rather than lectures. In the exhibitors' corner on the first floor, you could always meet many familiar faces from the industry.

(who)