I’m a postdoctoral researcher at the Data-Centric Computing group at Barcelona Supercomputing Center. I am currently the local PI of the EU Horizon OptimAIse project, where I work on improving the orchestration of modern AI systems on resource-constrained devices through energy-aware policies.

My research addresses the efficient and secure deployment of large-scale distributed AI systems. During my PhD, I have interned at IBM Research T.J. Watson, and IBM Research Ireland. Previously, I’ve done an academic exchange at DTU in Denmark, and I’ve been an Openlab Summer Student at CERN.

Selected publications:

• [CLOUD’25] Mind the Memory Gap: Unveiling GPU Bottlenecks in Large-Batch LLM Inference (arXiv)
  Pol G Recasens, Ferran Agullo, Yue Zhu, Chen Wang, Eun Kyung Lee, Olivier Tardieu, Jordi Torres, Josep Ll Berral

• [EuroMLSys’24] Towards Pareto Optimal Throughput in Small Language Model Serving (arXiv)
  Pol G Recasens, Yue Zhu, Chen Wang, Eun Kyung Lee, Olivier Tardieu, Alaa Youssef, Jordi Torres, Josep Ll Berral

• [NeurIPS’23] On masked pre-training and the marginal likelihood (arXiv)
  P Moreno-Muñoz, P Garcia Recasens, S Hauberg

In the past, I received a MSc in Data Science and BSc in Informatics Engineering from the Technical University of Catalonia (UPC). I also worked as a researcher for the Barcelona Supercomputing Center (BSC) and Denmark Technical University (DTU).

---

For further details, visit any of the following sections:

• Publications - updated publication list
• Teaching - teaching, talks and mentoring experience
• Work - past and current work experience
• Education - past education
• Contact - contact information

---