Hi, I’m Rafael!

I’m a first-year Computer Science PhD student at Carnegie Mellon University (CMU) and Instituto Superior Técnico (IST), supported by a CMU Portugal Dual Degree PhD fellowship. I’m co-advised by Profs. José Fragoso Santos, Limin Jia and Pedro Adão.

My research interests lie primarly in the areas of software security and formal methods. Most recently, I’ve been building robust analysis tools for detecting vulnerabilities in Node.js packages.

I hold an MSc in Computer Science and Engineering from Instituto Superior Técnico. I’ve also worked on fairness in Machine Learning as a Research Assistant at INESC-ID and served as a TA in multiple courses.

You can find my CV here.

PUBLICATIONS (Google Scholar)

• R. Gonçalves, F. Gouveia, I. Lynce, J. Fragoso Santos. Proxy Attribute Discovery in Machine Learning Datasets via Inductive Logic Programming. To appear in: TACAS 2025. [Preprint]

• R. Gonçalves, F. Ramos, P. Adão, J. Fragoso Santos. Poster: Specification-Driven Synthesis of Summaries for Symbolic Execution. Presented at: ISSTA/ECOOP 2023. [Poster]

Theses

• R. Gonçalves. Specification-Driven Synthesis of Summaries for Symbolic Execution. MSc thesis, Instituto Superior Técnico, 2023. [Thesis]

EXPERIENCE

Previously, I worked on the RIGA project as a Research Assisistant in the Automated Reasoning and Software Reliability group at at INESC-ID. Before that, I spent two months at CMU’s CyLab as a visiting scholar under the guidance of Prof. Limin Jia, working on NodeMedic.

Teaching

I’ve served as a TA in these courses at IST:

• Highly Dependable Systems (MSc) [Spring 2023, Spring 2024]
• Compilers (BSc) [Summer 2024]

PROJECTS

• RIGA: Reasoning Over Indirect Discrimination [DOI]
  The issue of fairness is a well-known challenge in Machine Learning (ML). Algorithmic bias can manifest during the training of ML models due to the presence of sensitive attributes, such as gender or racial identity. Project RIGA aims to apply automated reasoning techniques to detect indirect discrimination in ML classification algorithms. The focus is on so-called proxy attributes: those that, while not sensitive themselves, may lead to discrimiatory behavior due to their correlation with sensitive attributes.

CONTACT

Want to talk? Find me at:

GHC 9223, CMU

rgoncalv@andrew.cmu.edu

Rafael Gonçalves