About me
I am a doctoral student in the CMU-Portugal dual-degree PhD program in Software Engineering at Carnegie Mellon University (CMU) and Computer Science and Engineering at Instituto Superior Técnico (IST) - Universidade de Lisboa.
I am being kindly advised by Prof. Paolo Romano at IST and Prof. David Garlan at CMU. I am currently working as a graduate research assistant at Software and Societal Systems Department (S3D) where I integrate the ABLE Group, and as a researcher at IST and in the Distributed, Parallel and Secure Systems (DPPS) Group at INESC-ID Lisboa.
Research Interests
My research and main areas of interest are focused on Optimization, Machine Learning, Adversarial Training, Artificial Intelligence, Distributed Systems, Cloud Computing, Virtualization, and Computer Networks.
Publications
Information also available on Google Scholar.
Articles in Proceedings
Error-Driven Uncertainty Aware Training. Pedro Mendes, Paolo Romano, David Garlan. In ECAI 2024.
Hyper-parameter Tuning for Adversarially Robust Models. Pedro Mendes, Paolo Romano, David Garlan. In IJCAI-2024 AISafety Workshop 2024.
HyperJump: Accelerating HyperBand via Risk Modelling. Pedro Mendes, Maria Casimiro, Paolo Romano, David Garlan. In AAAI 2023.
TrimTuner: Efficient Optimization of Machine Learning Jobs in the Cloud via Sub-Sampling. Pedro Mendes, Maria Casimiro, Paolo Romano, David Garlan . In MASCOTS 2020.
Thesis
Leveraging Subsampling Techniques to Optimize Machine Learning Jobs in the Cloud. Pedro Mendes (supervised by Professors Paolo Romano and João Nuno Silva). MSc. Thesis, IST, Universidade de Lisboa, November 2019.
Other Articles
Exploring the Trade-offs to Train Robust Neural Models. Pedro Mendes, 2022.
Hyper-Parameter Tuning using Bayesian Optimization. Pedro Mendes, 2021.
Professinal Experience
Graduate Research Assistant 2021-OngoingSoftware and Societal Systems Department, Carnegie Mellon University, Pittsburgh (USA)
Researcher 2020-OngoingINESC-ID, Instituto Superior Técnico, Lisbon (Portugal)
Development of techniques to optimize the quality of Machine Learning models while reducing the training cost. Hyper-parameter tuning for training adversarially robust models. Improving the uncertainty estimation of neural models via uncertainty-aware training.
Research Intern at Priberam Dec 2023 - Jun 2024Lisbon (Portugal)
Hyper-parameter tuning for large language models Implementation of hyper-parameter tuning optimizers (e.g., HyperJump) on a popular Python framework called Optuna Development of the training pipeline to train language models for machine translation, summarization, and named entity recognition tasks
Junior Researcher Feb 2019 - Dec 2019INESC-ID, Instituto Superior Técnico, Lisbon (Portugal)
Optimization of the cloud resources and the hyper-parameters of Machine Learning models via subsampling techniques to enhance the efficiency of the model’s training while maximizing the models’ performance
Social
Contact