Diego Perez is now an Associate Professor at Linnaeus University in Växjö, Sweden. His research mainly focuses on how software can automatically assess and optimize its own behavior under uncertainty.
Here is what he had to say about their work and the future of performance engineering:
Could you briefly summarize the core contribution of your paper and what you are most proud of regarding its impact?
The paper established a taxonomy for classifying uncertainties in the runtime modeling of self-adaptive systems. While self-adaptive software relies on online model evaluations to navigate dynamic contexts, the runtime context information available to the system is rarely complete or accurate. Our work argued that uncertainty must be treated as a primary actor in model-based evaluation, classified distinct types of uncertainties that affect self-adaptive systems, and provided guidelines for managing them.
What I am most proud of is how this taxonomy helped the community by providing a structured vocabulary and framework incorporating runtime unknowns into model evaluations. Seeing researchers build upon this classification to treat uncertainty as a core modeling concern rather than an afterthought is deeply gratifying.
Looking back after ten years, did the impact or evolution of the work surprise you in any way? Is there anything you would approach differently if you were to do the work today?
I was pleasantly surprised by how quickly the treatment of uncertainty became a main focus in self-adaptive systems. If I were to approach the work today, I would place greater emphasis on characterizing and handling the types of uncertainty associated with data-driven machine learning techniques.
If you could send one message back to yourself while writing this paper, what would it be?
I would tell my past self to invest even more effort in expanding the specific concepts in the paper that proved to have the strongest long-term impact. Providing deeper theoretical and practical details on those core concepts would have offered even stronger guidance and reuse to researchers in the field over the following decade.
What exciting new challenges are you and your team focusing on at the moment?
Currently, my team and I are focusing on integrating sustainability concepts, such as carbon footprint metrics, into performance engineering, as well as exploring trustworthy autonomous decision-making under uncertainty stemming from AI component outputs, antifragile software architectures, and mechanisms for long-term self-evolution.
In your opinion, what emerging trends will dominate the field over the next few years?
Given the rapid pace at which the field evolves, accurately predicting future trends is inherently subject to its own uncertainty. However, looking at current trajectories, I see two interconnected open challenges:
First, performance engineering for AI modules will require focus from both AI providers to enhance underlying model performance and from developers of applications consuming AI services (such as LLMs or autonomous agents) to ensure acceptable end-to-end performance and predictability.
Second, we will need to manage the AI capability-performance-sustainability tradeoff. As AI services grow more sophisticated, performance engineering must balance the functional capabilities offered, the performance experienced by users, and the environmental sustainability (energy and carbon footprint) required to deliver them.
What is your top advice for junior researchers submitting to ICPE 2027?
I hope to attend ICPE 2027. My advice for junior researchers is to balance the excitement of exploring the unknown with grounding their problem in a realistic context. Exploratory or conceptual ideas gain immense strength when anchored in realistic operational constraints, showing researchers and practitioners that their novel technique works, that it matters in practice, and that they can readily adopt and reuse the proposed solutions.
What is one ICPE memory that has stayed with you?
ICPE was my primary venue and home conference across the formative years of my PhD and postdoctoral research. Beyond the technical presentations and keynotes, what stood out most was the warm, welcoming atmosphere where early career researchers could engage directly with senior figures in the field. Attending regularly throughout that crucial period of scholarly growth built a strong sense of community that has remained with me ever since.