Resource Aware Programming (2022-2023)
The goal of RAP is to explore how to provide developers immediate feedback about the impact of their programming in resources, such as execution time, memory and energy consumption. We are developing a type system that models these resources as probabilistic, and tracks them using symbolic execution à lá Liquid Types.
Funding: FCT, 2 years, 50K€
Advanced Computing (2021-2023) and (2022-2024)
Advanced Computing covers two Erasmus+ initiatives that I am coordinating at the University of Lisbon for the purpose of internationalisation of training in High Performance Computing and Artificial Intelligence.
Funding: Erasmus+, 880K€ for the consortium
CAMELOT was a large-scale CMU|Portugal project that aims to improve the efficiency and assurance of AI in fraud prediction. Our focus has been on interpretable AutoML, and assurance that machine learning pipelines have no bugs.
We developed an automatic approach to design domain-aware interpretable features, a type-guided genetic programming framework to embed domain knowledge, and we designed a verification tool to guarantee that machine learning pipelines have no semantic errors, such as data leakage, shuffling time-series, etc.
In this project, we automated the bioinformatics pipeline for producing diagnosis reports for medical genetic analysis. We also worked on understanding how to incorporate state-of-the-art prediction tools in clinical practice.
Funding: P2020, 625 k€