Profile picture

Rafael Pires

rafael.pires@epfl.ch +41 21 693 19 65 https://www.pires.tech/

EPFL IC IINFCOM SACS
BC 164 (Bâtiment BC)
Station 14
1015 Lausanne

Expertise

Computer systems, including challenges related to confidentiality, privacy, efficiency, and machine learning.
Rafael Pires is a lecturer and researcher at EPFL and holds a PhD in computer science (2019) from the University of Neuchâtel, Switzerland. His work focuses on systems-related challenges involving confidentiality, privacy, efficiency, and machine learning. He also holds two master’s degrees from Brazil, in computer science (2009) and mechatronics (2014), and has worked in both industry and the public sector, notably in digital forensics, embedded systems, robotics, and distributed systems.

Infoscience publications

Efficient Pyramidal Analysis of Gigapixel Images on a Decentralized Modest Computer Cluster

M. Reinbigler; R. Sharma; R. Pires; E. Brunet; A.-M. Kermarrec  et al.

2025. 31st International European Conference on Parallel and Distributed Computing (Euro-Par'25), Dresden, Germany, 2025-08-25 - 2025-08-29. p. 298 - 312. DOI : 10.1007/978-3-031-99872-0_21.

Detailed record

View at publisher

Low-Cost Privacy-Preserving Decentralized Learning

S. Biswas; D. Frey; A.-M. Kermarrec; Dimitri Lerévérend; F. Taïani  et al.

2025. The 25th Privacy Enhancing Technologies Symposium, Washington DC, USA, 2025-07-14 - 2025-07-19. p. 451 - 474. DOI : https://doi.org/10.56553/popets-2025-0108.

Detailed record

View at publisher

Leveraging Approximate Caching for Faster Retrieval-Augmented Generation

S. Bergman; Z. Ji; A.-M. Kermarrec; M. B. M. Randl; D. A. Petrescu  et al.

2025. 5th Workshop on Machine Learning and Systems (EuroMLSys), Rotterdam, The Netherlands, 2025-03-31. DOI : 10.1145/3721146.3721941.

Detailed record

View at publisher

Accelerating MoE Model Inference with Expert Sharding

O. Balmau; A.-M. Kermarrec; R. Pereira Pires; A. Loureiro Espírito Santo; M. A. de Vos  et al.

5th Workshop on Machine Learning and Systems (EuroMLSys), Rotterdam, The Netherlands, 2025-03-31.

Detailed record

View at publisher

Boosting Asynchronous Decentralized Learning with Model Fragmentation

S. Biswas; A.-M. Kermarrec; A. Marouani; R. Pereira Pires; R. Sharma  et al.

2025. The ACM Web Conference 2025, Sydney, Australia, 2025-04-28 - 2025-05-02.

Detailed record

Revisiting Ensembling in One-Shot Federated Learning

Y. Allouah; A. B. Dhasade; R. Guerraoui; N. Gupta; A.-M. Kermarrec  et al.

2024. 38th Annual Conference on Neural Information Processing Systems, Vancouver Convention Center, 2024-12-10 - 2024-12-15.

Detailed record

Noiseless Privacy-Preserving Decentralized Learning

S. Biswas; M. Even; L. Massoulié; A.-M. Kermarrec; R. Pereira Pires  et al.

2024. The 25th Privacy Enhancing Technologies Symposium, Washington DC, USA, 2025-07-14 - 2025-07-19. p. 824 - 844. DOI : 10.56553/popets-2025-0043.

Detailed record

View at publisher

Energy-Aware Decentralized Learning with Intermittent Model Training

M. A. de Vos; A. B. Dhasade; P. Dini; E. Guerra; M. Miozzo  et al.

38th IEEE International Parallel & Distributed Processing Symposium, San Francisco, California, USA, 2024-05-27 - 2024-05-31.

Detailed record

View at publisher

Epidemic Learning: Boosting Decentralized Learning with Randomized Communication

M. A. de Vos; S. Farhadkhani; R. Guerraoui; A.-M. Kermarrec; R. Pereira Pires  et al.

2023. 37th Conference on Neural Information Processing Systems (NeurIPS 2023), New Orleans, Louisiana, USA, December 10-16, 2023.

Detailed record

Decentralized learning made easy with DecentralizePy

A. B. Dhasade; A.-M. Kermarrec; R. Pereira Pires; R. Sharma; M. Vujasinovic

2023. 3rd Workshop on Machine Learning and Systems (EuroMLSys'23), Rome, Italy, May 8th. DOI : 10.1145/3578356.3592587.

Detailed record

View at publisher

Malware in the SGX supply chain: Be careful when signing enclaves!

V. Craciun; P. Felber; A. Mogage; E. Onica; R. Pereira Pires

IEEE Transactions on Dependable and Secure Computing. 2022. DOI : 10.1109/TDSC.2020.3024562.

Detailed record

View at publisher

PhD Students

Mathis Benjamin Manuel Randl

Courses

Information, Computation, Communication

CS-119(g)

The objective of this course is to introduce students to algorithmic thinking, to familiarize them with the fundamentals of computer and communications sciences and to develop their first programming skills (in C++).

Information, Computation, Communication

CS-119(k)

On one side, this course covers the concepts of algorithms, the representation of information, signal sampling and compression, and an overview of systems (CPU, memory, etc.). On the other side, an introduction to programming in Python is given.

Introduction to programming

CS-107

This course presents the fundamentals of programming and object-oriented programming (using the Java language). It also gives an introduction to a computer development environment (by default on Linux).

Project oriented programming

COM-112(a)

This course focuses on the complementary features of the C++ language that allow to design robust modular applications (principle of separation of concerns). The practice dimension is deemed particularly important ; for this reason a significant time is devoted to a project.