Sabela Ramos
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sabela Ramos has a total of 109 co-authors. Here are the top co-authors by number of shared publications:
- Guillermo L. Taboada
Computer Architecture Group, CITIC, Universidade da Coruña, Campus de Elviña, 15071 A Coruña, Spain
11 shared publications - Juan Touriño
Universidade da Coruña, CITIC, Computer Architecture Group, A Coruña, Spain
11 shared publications - Ramón Doallo
University of A Coruña, Spain
10 shared publications - Roberto R. Exposito
Universidade da Coruña, CITIC, Computer Architecture Group, A Coruña, Spain
9 shared publications Léonard Hussenot
Google, Mountain View, USA
2 shared publicationsTorsten Hoefler
Scalable Parallel Computing Lab, Computer Science Department, ETH Zürich, Switzerland
2 shared publications
Latest Publications
- Open Access
- 6 October 2026
- Published by Springer Nature in Nature Medicine
- Vol. 32 (10) , 3695-3704
An open vision-language model for diverse medical applications
Artificial intelligence has high potential for impact in healthcare applications, but its training and deployment are challenging due to diverse data, a complex spectrum of possible tasks and important privacy needs. High-performing foundation models that enable data-efficient fine-tuning for diverse downstream tasks can meaningfully accelerate development in this domain. We introduce MedGemma, a collection of medical vision-language foundation models based on Gemma 3. MedGemma demonstrates advanced medical understanding and reasoning across images and text and multiple medical imaging domains, exceeding the performance of similarly sized generative models while maintaining the general capabilities of the Gemma base models. For out-of-distribution tasks, MedGemma achieves improvements of 2.6–10% in medical image question answering, 15.5–18.1% in chest X-ray finding classification and 10.8% in agentic evaluations compared with the base models. Our results show that fine-tuning MedGemma can be more effective than fine-tuning the base Gemma 3 model for medical tasks, particularly in the setting of limited training data. We additionally introduce MedSigLIP, a medically tuned vision encoder derived from SigLIP. MedSigLIP powers the visual understanding capabilities of MedGemma and, as an encoder, achieves performance comparable to or better than that of many specialized medical image encoders. Taken together, the MedGemma collection provides a strong foundation of medical image and text capabilities, with the potential to accelerate medical research and the development of downstream applications.
- 6 October 2026
- Published by Springer Nature in Nature Medicine
- Vol. 32 (10) , 3695-3704
Open AccessAn open vision-language model for diverse medical applications
Artificial intelligence has high potential for impact in healthcare applications, but its training and deployment are challenging due to diverse data, a complex spectrum of possible tasks and important privacy needs. High-performing foundation models that enable data-efficient fine-tuning for diverse downstream tasks can meaningfully accelerate development in this domain. We introduce MedGemma, a collection of medical vision-language foundation models based on Gemma 3. MedGemma demonstrates advanced medical understanding and reasoning across images and text and multiple medical imaging domains, exceeding the performance of similarly sized generative models while maintaining the general capabilities of the Gemma base models. For out-of-distribution tasks, MedGemma achieves improvements of 2.6–10% in medical image question answering, 15.5–18.1% in chest X-ray finding classification and 10.8% in agentic evaluations compared with the base models. Our results show that fine-tuning MedGemma can be more effective than fine-tuning the base Gemma 3 model for medical tasks, particularly in the setting of limited training data. We additionally introduce MedSigLIP, a medically tuned vision encoder derived from SigLIP. MedSigLIP powers the visual understanding capabilities of MedGemma and, as an encoder, achieves performance comparable to or better than that of many specialized medical image encoders. Taken together, the MedGemma collection provides a strong foundation of medical image and text capabilities, with the potential to accelerate medical research and the development of downstream applications.
- Open Access
RLDS: an Ecosystem to Generate, Share and Use Datasets in Reinforcement Learning
We introduce RLDS (Reinforcement Learning Datasets), an ecosystem for recording, replaying, manipulating, annotating and sharing data in the context of Sequential Decision Making (SDM) including Reinforcement Learning (RL), Learning from Demonstrations, Offline RL or Imitation Learning. RLDS enables not only reproducibility of existing research and easy generation of new datasets, but also accelerates novel research. By providing a standard and lossless format of datasets it enables to quickly test new algorithms on a wider range of tasks. The RLDS ecosystem makes it easy to share datasets without any loss of information and to be agnostic to the underlying original format when applying various data processing pipelines to large collections of datasets. Besides, RLDS provides tools for collecting data generated by either synthetic agents or humans, as well as for inspecting and manipulating the collected data. Ultimately, integration with TFDS facilitates the sharing of RL datasets with the research community.
Open AccessRLDS: an Ecosystem to Generate, Share and Use Datasets in Reinforcement Learning
We introduce RLDS (Reinforcement Learning Datasets), an ecosystem for recording, replaying, manipulating, annotating and sharing data in the context of Sequential Decision Making (SDM) including Reinforcement Learning (RL), Learning from Demonstrations, Offline RL or Imitation Learning. RLDS enables not only reproducibility of existing research and easy generation of new datasets, but also accelerates novel research. By providing a standard and lossless format of datasets it enables to quickly test new algorithms on a wider range of tasks. The RLDS ecosystem makes it easy to share datasets without any loss of information and to be agnostic to the underlying original format when applying various data processing pipelines to large collections of datasets. Besides, RLDS provides tools for collecting data generated by either synthetic agents or humans, as well as for inspecting and manipulating the collected data. Ultimately, integration with TFDS facilitates the sharing of RL datasets with the research community.
- 11 January 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 27 (10) , 2824-2837
Cache Line Aware Algorithm Design for Cache-Coherent Architectures
The increase in the number of cores per processor and the complexity of memory hierarchies make cache coherence key for programmability of current shared memory systems. However, ignoring its detailed architectural characteristics can harm performance significantly. In order to assist performance-centric programming, we propose a methodology to allow semi-automatic performance tuning with the systematic translation from an algorithm to an analytic performance model for cache line transfers. For this, we design a simple interface for cache line aware optimization, a translation methodology, and a full performance model that exposes the block-based design of caches to middleware designers. We investigate two different architectures to show the applicability of our techniques and methods: the many-core accelerator Intel Xeon Phi and a multi-core processor with a NUMA configuration (Intel Sandy Bridge). We use mathematical optimization techniques to tune synchronization algorithms to the microarchitectures, identifying three techniques to design and optimize data transfers in our model: single-use, single-step broadcast, and private cache lines.
- 11 January 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 27 (10) , 2824-2837
Cache Line Aware Algorithm Design for Cache-Coherent Architectures
The increase in the number of cores per processor and the complexity of memory hierarchies make cache coherence key for programmability of current shared memory systems. However, ignoring its detailed architectural characteristics can harm performance significantly. In order to assist performance-centric programming, we propose a methodology to allow semi-automatic performance tuning with the systematic translation from an algorithm to an analytic performance model for cache line transfers. For this, we design a simple interface for cache line aware optimization, a translation methodology, and a full performance model that exposes the block-based design of caches to middleware designers. We investigate two different architectures to show the applicability of our techniques and methods: the many-core accelerator Intel Xeon Phi and a multi-core processor with a NUMA configuration (Intel Sandy Bridge). We use mathematical optimization techniques to tune synchronization algorithms to the microarchitectures, identifying three techniques to design and optimize data transfers in our model: single-use, single-step broadcast, and private cache lines.
- 5 July 2014
- Published by Springer Nature in Computer Science - Research and Development
- Vol. 30 (2) , 207-218
Exploring energy-performance-quality tradeoffs for scientific workflows with in-situ data analyses
- 5 July 2014
- Published by Springer Nature in Computer Science - Research and Development
- Vol. 30 (2) , 207-218
Exploring energy-performance-quality tradeoffs for scientific workflows with in-situ data analyses
- 10 April 2014
- Published by Wiley in Concurrency and Computation: Practice and Experience
- Vol. 27 (5) , 1169-1187
Nonblocking collectives for scalable Java communications
Summary: This paper presents a Java implementation of the recently published MPI 3.0 nonblocking message passing collectives in order to analyze and assess the feasibility of taking advantage of these operations in shared memory systems using Java. Nonblocking collectives aim to exploit the overlapping between computation and communication for collective operations to increase scalability of message passing codes, as it has been carried out for nonblocking point‐to‐point primitives. This scalability has become crucial not only for clusters but also for shared memory systems because of the current trend of increasing the number of cores per chip, which is leading to the generalization of multi‐core and many‐core processors. Message passing libraries based on remote direct memory access, thread‐based progression, or implementing pure multi‐threading shared memory support could potentially benefit from the lack of imposed synchronization by nonblocking collectives. But, although the distributed memory scenario has been well studied, the shared memory one has not been tackled yet. Hence, nonblocking collectives support has been included in FastMPJ, a Message Passing in Java (MPJ) implementation, and evaluated on a representative shared memory system, obtaining significant improvements because of overlapping and lack of implicit synchronization, and with barely any overhead imposed over common blocking operations. Copyright © 2014 John Wiley & Sons, Ltd.
- 10 April 2014
- Published by Wiley in Concurrency and Computation: Practice and Experience
- Vol. 27 (5) , 1169-1187
Nonblocking collectives for scalable Java communications
Summary: This paper presents a Java implementation of the recently published MPI 3.0 nonblocking message passing collectives in order to analyze and assess the feasibility of taking advantage of these operations in shared memory systems using Java. Nonblocking collectives aim to exploit the overlapping between computation and communication for collective operations to increase scalability of message passing codes, as it has been carried out for nonblocking point‐to‐point primitives. This scalability has become crucial not only for clusters but also for shared memory systems because of the current trend of increasing the number of cores per chip, which is leading to the generalization of multi‐core and many‐core processors. Message passing libraries based on remote direct memory access, thread‐based progression, or implementing pure multi‐threading shared memory support could potentially benefit from the lack of imposed synchronization by nonblocking collectives. But, although the distributed memory scenario has been well studied, the shared memory one has not been tackled yet. Hence, nonblocking collectives support has been included in FastMPJ, a Message Passing in Java (MPJ) implementation, and evaluated on a representative shared memory system, obtaining significant improvements because of overlapping and lack of implicit synchronization, and with barely any overhead imposed over common blocking operations. Copyright © 2014 John Wiley & Sons, Ltd.
- 6 February 2014
- Published by Springer Nature in Cluster Computing
- Vol. 17 (3) , 1031-1050
FastMPJ: a scalable and efficient Java message-passing library
- 6 February 2014
- Published by Springer Nature in Cluster Computing
- Vol. 17 (3) , 1031-1050
FastMPJ: a scalable and efficient Java message-passing library
- Open Access
- 17 June 2013
- Published by Association for Computing Machinery (ACM)
- p. 97-108
Modeling communication in cache-coherent SMP systems
- 17 June 2013
- Published by Association for Computing Machinery (ACM)
- p. 97-108
Open AccessModeling communication in cache-coherent SMP systems
- 1 May 2013
- Published by Elsevier in Science of Computer Programming
- Vol. 78 (5) , 425-444
Java in the High Performance Computing arena: Research, practice and experience
- 1 May 2013
- Published by Elsevier in Science of Computer Programming
- Vol. 78 (5) , 425-444
Java in the High Performance Computing arena: Research, practice and experience
- 1 March 2013
- Published by Springer Nature in Journal of Grid Computing
- Vol. 11 (4) , 613-631
Analysis of I/O Performance on an Amazon EC2 Cluster Compute and High I/O Platform
- 1 March 2013
- Published by Springer Nature in Journal of Grid Computing
- Vol. 11 (4) , 613-631
Analysis of I/O Performance on an Amazon EC2 Cluster Compute and High I/O Platform
- 1 March 2013
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1398-1404
Evaluation of Java for General Purpose GPU Computing
The presence of many-core units as accelerators has been increasing due to their ability to improve the performance of highly parallel workloads. General Purpose GPU(GPGPU) computing has allowed the graphical units to emerge as successful co-processors that can be employed to improve the performance of many different non-graphical applications with high parallel requirements, which make them suitable for many High Performance Computing workloads. While the main libraries developed to exploit the massive parallel capacity of GPUs are oriented to C/C++ programmers, there have been several efforts to extend this support to other languages. Among them, Java stands out for being one of the most extended languages and there are multiple projects that try to enable Java to take advantage of GPGPU computing. In this scenario, this paper presents an evaluation of the most relevant among the current solutions that exploit GPGPU computing in Java.
- 1 March 2013
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1398-1404
Evaluation of Java for General Purpose GPU Computing
The presence of many-core units as accelerators has been increasing due to their ability to improve the performance of highly parallel workloads. General Purpose GPU(GPGPU) computing has allowed the graphical units to emerge as successful co-processors that can be employed to improve the performance of many different non-graphical applications with high parallel requirements, which make them suitable for many High Performance Computing workloads. While the main libraries developed to exploit the massive parallel capacity of GPUs are oriented to C/C++ programmers, there have been several efforts to extend this support to other languages. Among them, Java stands out for being one of the most extended languages and there are multiple projects that try to enable Java to take advantage of GPGPU computing. In this scenario, this paper presents an evaluation of the most relevant among the current solutions that exploit GPGPU computing in Java.