Generative and Agentic AI will radically transform traditional software systems from static, rigidly composed architectures into dynamic & adaptive cognitive organisms. In recent work, we advocated for a paradigm shift for building future GenAI-native systems, and presented a holistic approach to design such systems (see https://doi.org/10.1145/3759429.3762620).

Autonomous composability and evolvability are central to the success of such GenAI-native systems. However, achieving these capabilities introduces significant challenges, including ensuring overall stability, assurance, and efficiency in such complex, organic systems. Addressing these challenges demands the development of novel design principles, abstractions, formalisms, and foundational frameworks to enable and manage composability effectively.

Key objectives of the internship:

• Analyze and identify the key challenges associated with autonomous composability and evolvability in GenAI-native systems.

• Evaluate the limitations of existing techniques in addressing these challenges.

• Innovate and propose novel design principles, formalisms, and/or enabling frameworks to enhance autonomous composability and adaptability in GenAI-native systems.

This internship offers an opportunity to contribute to cutting-edge research and shape the future of GenAI-native software systems.

• Student enrolled in Ph.D. Computer Science/Engineering in software engineering, AI and/or autonomous systems

• Strong programming skills in Python

• Language skills: English

• Experience in AI agents, autonomous systems and/or software engineering

• A strong publication record is a big plus.

• You will review scientific literature on this topic

• You will design novel techniques for autonomously composing GenAI-native systems, and compare these with SotA techniques on multi-agent orchestration and communication

• You will evaluate the effectiveness of the proposed techniques on representative problems and use cases, and study their limitations.

• Consolidate your research in a scientific publication.

Location: Antwerp (Belgium)