About Me
I am a machine learning researcher working with Professor Paolo Romano at the Distributed, Parallel and Secure Systems group of INESC-ID.
I recently received a dual-degree PhD in Software Engineering by Carnegie Mellon University and Instituto Superior Técnico, University of Lisbon. I was kindly advised by professors David Garlan, from the ABLE group, Software and Society Systems Department (S3D), Carnegie Mellon University and Paolo Romano from INESC-ID and Instituto Superior Técnico, University of Lisbon.
While pursuing my PhD, I concluded and received a MSc. in Software Engineering from Carnegie Mellon University (2022). In 2016 and 2018 I received my BSc. and (respectively) in Electrical and Computer Engineering, from Instituto Superior Técnico, with a major in Computers and a minor in Systems, Decision and Control.
Research interests
Self-Adaptive ML-based systems and Sustainable ML
Throughout my PhD I contributed to the state-of-the-art in these
areas by engineering self-adaptive ML-based systems that automatically determine
whether their ML models should be retrained/fine-tuned by analyzing the cost/benefits
trade-offs of doing so. Our results on fraud detection and
machine translation systems demonstrated that (i) such self-adaptive
systems are more sustainable than systems that simply update the ML models periodically
(i.e., with fixed periodicities) and (ii) quality of service improves when compared against
both non-adaptive and periodic baselines.
Hyper-parameter Optimization
Together with a fellow PhD student, we created TrimTuner
and HyperJump, two systems that perform hyper-parameter optimization
and improved over the state-of-the-art. Additionally, I often review for top ML conferences
such as ICML, NeurIPS, and AAAI.
Publications
Information also available from my Google Scholar profile.
-
FLEXICO: Sustainable Machine Translation via Self-Adaptation
SEAMS 2025 - 20th International Conference on Software Engineering for Adaptive and Self-Managing Systems
-
Self-Adapting Machine Learning-based Systems via a Probabilistic Model Checking Framework
TAAS 2024 - ACM Transactions on Autonomous and Adaptive Systems
-
HyperJump: Accelerating HyperBand via Risk Modelling
AAAI 2023 - Proceedings of the AAAI Conference on Artificial Intelligence
-
Towards a Framework for Adapting Machine Learning Components
ACSOS 2022 - IEEE International Conference on Autonomic Computing and Self-Organizing Systems
-
A Probabilistic Model Checking Approach to Self-Adapting Machine Learning Systems
ASYDE 2021 - 3rd International Workshop on Automated and verifiable Software sYstem DEvelopment, Co-located with SEFM 2021
-
Self-adaptive Machine Learning Systems: Research Challenges and Opportunities
-
Self-Adaptation for Machine Learning Based Systems
SAML 2021 - 1st International Workshop on Software Architecture and Machine Learning, Co-located with ECSA
-
TrimTuner: Efficient Optimization of Machine Learning Jobs in the Cloud via Sub-Sampling
MASCOTS 2020 - IEEE Symposium on Modelling, Analysis, and Simulation of Computer and Telecommunication Systems
-
A Quest for Inspiration: How Users Create and Reuse PINs
WAY 2020 - Who Are You?! Adventures in Authentication Workshop
-
Lynceus: Cost-efficient Tuning and Provisioning of Data Analytic Jobs
ICDCS 2020 - 40th IEEE International Conference on Distributed Computing Systems
-
Reduzindo o Custo de Explorar Configurações para Execução de Aplicações na Nuvem
Education
-
Dual PhD in Software Engineering / Computer Science
2019 - 2024
Thesis Self-Adaptive Machine Learning-based Systems
Carnegie Mellon University & IST, University of Lisbon (pass with distinction)
-
MSc. Thesis Electrical and Computer Engineering
2016 - 2018
Thesis Lynceus: Long-Sighted, Budget-Aware Online Tunning of Cloud Applications
Final course average: 17/20 (A ECTS)
IST, University of Lisbon, November 2018
Experience
-
Research collaboration
May 2022 - Aug 2024- Created a self-adaptive machine translation system that automatically determines whether to fine-tune the machine translation model.
-
Research intern
May 2021 - July 2021- Worked on the problem of real-time performance monitoring for credit/debit card fraud detection systems. Specifically, on how to evaluate the quality of the fraud detetcion model in the absence of ground truth labels for the credit/debit card transactions.
- Related article on AI Observability.
-
Teaching assistant
August 2022 - December 2022
- 17-614 Formal Methods
- 17-624 Advanced Formal Methods
May 2022 - July 2022
- IAC: Introduction to computer architectures
-
DPSS Researcher
September 2017 - Present-
CRAI
enhance the efficiency and scalability of AI models via: system-level optimization,
cloud-level optimization and self-adaptation. -
CAMELOT
improve the efficiency, accuracy, and/or quality of Feedzai's fraud detection
machine learning platform and associated stack using efficient cloud resources.
-
CRAI