On 16 November at 15:15, DEI welcomes Francisco Ortega, Associate Professor at Colorado State University, as a guest speaker for the DEI Talks. In room I-105, in a session open to all and moderated by Prof. Rui Rodrigues (DEI), he will present “XR Visual Search and Notifications”.

About the DEI Talk:

“Augmented reality (AR) can put an arrow on the object you are looking for. The questions are where to place that arrow and whether the cue is correct — and what information-access effort costs. Related to visual search cues, we have notifications and timers: how do they compete for attention with the world they exist to explain, and where do we place them? Our work in visual search and information-access effort includes timers, cues, and notifications. This line of research has several applications, from civilian use cases (e.g., rescue) and health applications (e.g., CPR) to military applications (e.g., searching for targets).”

About the Speaker:

“Francisco Ortega is an Associate Professor of Computer Science at Colorado State University and director of the Natural User Interaction Laboratory. His research builds and studies multimodal interaction (gesture, speech, gaze, and touch) for augmented and virtual reality, with a strong commitment to accessibility and to training the next generation of computing researchers. Before joining CSU, he completed his Ph.D. and a postdoc at Florida International University, where the lab had its first home. His work has been supported by the NSF (including a CAREER award), the Office of Naval Research, DARPA, and the Dan Marino Foundation, among others. He has authored three books on 3D user interfaces, elicitation methodology, and multimodal interaction, and holds three U.S. patents.”

The 21st Creativity Talk welcomes Philippe Pasquier, Professor at the School of Interactive Arts and Technology (SIAT) at Simon Fraser University, as guest speaker. On 3 December at 17:30, he will present “Alternative Approaches to Creative AI: Diversity, Literacy, and Depolarisation”, a talk on ethical, small-data alternatives to the dominant generative tools, and on how they can give control back to artists. The session will be moderated by Prof. Gilberto Bernardes (DEI) and will be streamed on the Creativity Talks YouTube channel, here.

On 16 November at 14:00, DEI welcomes Bruno Magalhães, Principal Researcher at Huawei Research Zurich, as a guest of the DEI Talks. In room I-105, in a session open to all, he will present “Towards efficient AI models: MegaKernels for distributed training and efficient inference of large AI models”.

About the DEI Talk:

“The scale of today’s AI models, with billions of parameters, turns their training and inference into distributed computing problems, bound mainly by data movement between memory, compute units and the network. Training needs compute large storage to hold billions of training parameters, and large compute power (FLOPs) to train the model on massive datasets in a feasible time. Inference requires low runtime for a positive user experience, and low running costs (“dollar per FLOP”). This talk will cover it.

For distributed training, the talk presents the main parallelization methods and the trade-off each one makes between memory per GPU and communication volume, given the network topology: data parallelism with sharding of the model state (ZeRO, FSDP), pipeline parallelism (1F1B, Zero Bubble, DualPipe), tensor parallelism (Megatron-LM), sequence and context parallelism (Ulysses, Ring Attention) and expert parallelism.

For efficient inference, we’ll discuss the roofline model that separates the compute-bound prefill from the memory-bandwidth-bound decode. Then techniques such as KV caching, PagedAttention, continuous batching, Grouped-Query Attention, Multi-head Latent Attention.

We will then introduce the concept of MegaKernels, adopted by systems such as DeepSeek to maximize performance. These kernels fuse computation and inter-GPU communication into a single kernel, reduce memory bandwidth, remove kernel launch runtime, and allows for computation/communication overlap. We will detail real implementations on the use cases of Tensor Parallelism, Expert Sharding and Expert parallelism.

Finally we will show that network latency (bandwidth) is critical to large model inference (training), and demonstrate how distributed model performance can be further optimized by programable switches that perform in-network computing of AI-related operations. We briefly introduce a performance modeling technique that extrapolates model performance onto new (unseen) network environment.

The methods are illustrated on Transformers, Mixture-of-Experts (MoE) models, diffusion and agentic models.”

About the Speaker:

“Bruno Magalhães is a principal AI Systems and Networks at Huawei Research, focusing on the efficiency for AI MegaKernels. Before he was a researcher at Microsoft Research where he worked on ML models for storage and reading of data in glass (Project Silica). Bruno holds a PhD from EPFL in Computational Neuroscience where he focused on (distributed) variable timestep interpolation of the electrical activity of the brain.”

On 7 October, Tiago Azevedo, a Research Associate at the Department of Computer Science and Technology and Research Fellow at Hughes Hall, University of Cambridge, will be the guest speaker of the DEI Talks series, presenting “Graph-Based Approaches to Biological Systems: Applications in Neuroimaging and Molecular Networks.” The session will take place in room B006 at 14:30 and will be moderated by Professor Rosaldo Rossetti (DEI).

An alumnus of DEI-FEUP, Tiago completed his Integrated Master’s Degree in Informatics and Computing Engineering in 2015 under the supervision of Professor Rosaldo Rossetti. Following his studies in Porto, he continued his academic journey at the University of Cambridge, where he earned his PhD and currently conducts his research. This will be an excellent opportunity to learn about the career path of a former DEI student who is now contributing to advances in artificial intelligence applied to some of the most significant challenges in health and biology.

About the DEI Talk:

“Biological systems can be represented as networks in which nodes correspond to brain regions, genes, or proteins, and edges capture functional, statistical, or potentially causal relationships. In this talk, I will discuss how graph-based methods can be used to model such systems, focusing first on a deep graph neural network architecture for learning spatio-temporal dynamics from resting-state functional MRI data.
I will then turn to molecular networks, briefly discussing previous work on multilayer modelling of the human transcriptome (i.e., gene expression). The main focus will then be my current work on causal discovery and stability analysis in high-dimensional proteomic data, motivated by Parkinson’s disease. Specifically, I will describe how traditional causal graph estimation and consensus analysis can be used to identify molecular relationships and communities that are robust to variation in the data.”

About the Speaker:

“Tiago Azevedo is a research associate at the Department of Computer Science and Technology and a research fellow at Hughes Hall, at the University of Cambridge, where he broadly likes to research applications of artificial intelligence in the medical and biological domains.
His current position, funded under the Sustaining Innovation Postdocs (SIPD) programme by Astex Pharmaceuticals, centres on applying machine learning to better understand Parkinson’s disease. Specifically, uncovering biologically grounded patient subtypes by integrating multi-modal data. Prior to his current focus on Parkinson’s disease, Tiago worked on cancer models using Raman spectroscopy histology data as part of CHARM, one of the first ever European Innovation Council Transition Challenges projects funded by the EU. He also spent over a year as a senior research scientist at Arm, working on efficient probabilistic machine learning and graph neural networks.”

The twentieth session of the Creativity Talks will feature Denise de Souza Fleith, Full Professor and Senior Researcher at the Institute of Psychology of the University of Brasília, as the keynote speaker.

The session will focus on “Creativity in Education: Are We Prepared for the Future?”. The presentation will be delivered in Portuguese and will take place online on October 8, 2026, at 17:30, moderated by Manuel Firmino Torres, Assistant Professor at FEUP’s Department of Informatics Engineering (DEI).

Learning to think and act creatively has become a fundamental requirement for personal, academic, and professional development in contemporary society. To address the challenges of the 21st century and prepare new generations for an increasingly complex and unpredictable future, it is essential that students see themselves as active participants in their own learning and believe in their capacity to contribute to collective well-being and the construction of a fairer and more sustainable society. This presentation aims to discuss the role of creativity in educational contexts in light of theoretical models, empirical evidence, and pedagogical practices grounded in scientific research. It seeks to examine the relationship between creativity and teaching and learning processes, identifying factors that foster or inhibit the expression of creative potential, as well as pedagogical practices and classroom climate characteristics that promote creativity among students and teachers. It is important to emphasize that creativity is not limited to inspiration or an innate gift; rather, it is a complex and dynamic phenomenon resulting from the interaction of personal, social, cultural, and historical factors. From this perspective, promoting creativity in education involves creating learning environments that value curiosity, autonomy, critical thinking, collaboration, and creative problem-solving, thereby better preparing individuals for the challenges of both the present and the future.

About the Speaker:

“Denise de Souza Fleith is a psychologist and holds a PhD in Educational Psychology from the University of Connecticut. She completed postdoctoral studies at the National Academy for Gifted and Talented Youth in England and at the University of Minho in Portugal. She is a Full Professor and Senior Researcher at the Institute of Psychology of the University of Brasília, a researcher with the National Council for Scientific and Technological Development (CNPq), and a member of the Scientific and Technical Committee of the National Science for Education Network. She is a Fellow of the International Society for the Study of Creativity and Innovation and served as President of the World Council for Gifted and Talented Children (2021-2025).”

Watch on the CT’s YouTube channel: https://www.youtube.com/watch?v=f30JR_nHrDA

Schedule: Google Calendar

The nineteenth session of the Creativity Talks took place on September 21, 2026, at the Faculty of Engineering of the University of Porto (FEUP) and featured Daniela Damian, Professor and ECS-CAPI Chair in Inclusive Science, Technology and Engineering (Lero Parnas Scholar) in the Department of Computer Science at the University of Victoria, as the keynote speaker. The session, moderated by Rui Maranhão, Full Professor at DEI, focused on the topic “The Human Connection: The Essential Compass to Innovating Technologies That Matter”.

In her presentation, Daniela Damian reflected on the growing importance of understanding users and their experiences in a world increasingly shaped by digital technologies and artificial intelligence. She highlighted that, although user-centered approaches are widely recognized in software engineering, human connection remains essential for developing inclusive technologies with meaningful impact. Drawing on several software engineering projects addressing societal challenges, she emphasized the importance of skills such as relationship-building, authentic empathy, perspective-taking, and cultural sensitivity, arguing that these competencies are fundamental to ensuring that technological innovation remains guided by dignity, equity, and human needs.

About the Speaker

Daniela Damian is a Professor of Software Engineering and the ECS-CAPI Chair in Inclusive Science, Technology, and Engineering at the University of Victoria. Over the years, together with her students, she has led empirical research on the human and social aspects of requirements engineering and software development, with a particular focus on teamwork, remote communication, and collaboration. She is the founder and Director of the INSPIRE:STEM for Social Impact Program at the University of Victoria, where diverse teams engage in community-driven, experiential-learning science and engineering projects with societal impact.

Daniela Damian has served as Program Chair for major Software Engineering conferences and was, among others, Co-Chair of the inaugural Conference on Software Engineering in Society (2014), RE 2019, and ICSE 2022. She has served on the editorial boards of IEEE Transactions on Software Engineering and currently serves on the Advisory Board of the Empirical Software Engineering Journal.

Her achievements have been recognized through several awards, including the 2019 Faculty of Engineering Teaching Award and the 2020 Provost’s REACH Award for Excellence in Teaching for Experiential Learning. In 2019, she received the Royal Society of New Zealand Catalyst: International Leader Award, and in 2024, she became a Lero Parnas Fellow. An ACM Distinguished Speaker, she was recently honored with the 2026 SIGSOFT Influential Educator Award.

Further information about her work and research can be found at danieladamian.ca, including her TEDxVictoria talk on the role of empathy in designing technology for a more diverse and compassionate world.