Join the division and AI Engineering Lab to be at the forefront of research in agentic AI software systems. This PhD project aims to observe and monitor agentic AI systems using intentional visual models – models where intentions or goals are explicit and evaluated during system operations. The PhD student will be supervised by Assoc. Prof. Jennifer Horkoff and will be part of the Interaction Design and Software Engineering Division of the Department of Computer Science and Engineering.
The Department of Computer Science and Engineering, a joint department of Chalmers and the University of Gothenburg, spans the breadth of computing disciplines. Our internationally visible research, strong industry links and diverse environment create a collaborative setting where ideas grow into real impact.
At the division of Interaction Design and Software Engineering, we design smarter ways to engineer better software, and explore how people engage with digital systems, combining global research perspectives with strong collaboration with industry.
The PhD position will be part of the Chalmers and University of Gothenburg’s AI Engineering Lab, focusing on practices, knowledge, and experience making AI advancements work in real and complex software systems. The PhD student will part of the WASP graduate school.
Project description
Agentic AI systems allow multiple foundation-model powered agents to work together, often using external tools, to accomplish system objectives. AI agent configurations can be complicated and dynamic, with many possible configurations of agents, agent types, tasks, inputs and outputs, including human oversight, memory, and mechanisms for reflection and improvement. Thus, there is a need to observe the status of the system in terms of current agent roles and tasks, goals (intentions), inputs and outputs, in order to evaluate whether the goals of individual agents and the overall agentic system are satisfied, including whether improvement mechanisms are effective. The overall objective of this PhD proposal is to support automated and intentional visualization, observability, and monitoring of agentic AI systems. We propose to use and adapt existing and well-established work in agent- and goal-oriented modeling frameworks for this purpose.