Andrea Piras

Andrea Piras

Causal Inference and Graph-based Learning.

Andrea Piras - Personal Academic Page

About me

Illustration of a network 👨🏻‍💻 I’m a second year Ph.D. student in the Department of Computer Science at the University of Illinois Chicago (UIC).

  • I am working under the supervision of professor Elena Zheleva in the EDGES lab, a statistical relational learning and data science lab.

🔬 My research interests lie at the intersection of Causal Inference, Data Science, and Graph-based Learning

📚 I’m currently working on causal modeling in network data

🎓 Education

I hold a double Master’s degree in Computer Science, awarded with Summa Cum Laude jointly by the University of Illinois Chicago and Politecnico di Milano.

  • At UIC, I graduated with a 4.0 GPA, taking advanced coursework in Machine Learning, Causal Inference, and NLP.
  • At Politecnico, I graduated with a final mark of 110L/100 (GPA of 3.82), specializing in Data Mining and Software Engineering.

I also hold a Bachelor’s degree in Engineering of Computing Systems from Politecnico di Milano, where I graduated with full marks (110/110).

🔬 Research Experience

Research at UIC

In 2024, I served as a Graduate Hourly Assistant at the University of Illinois Chicago, where I developed algorithms for causal discovery in relational settings using graphical models.

Research at NECSTLab

From 2020 to 2024, I was a Research Scholar at the NECSTLab research laboratory at Politecnico di Milano. I contributed to multiple projects focused on graph algorithms, genome graph analysis, and infrastructure modeling.

Highlighted projects include:

  • CPIExtract: for my Master’s thesis, I conducted research in collaboration with Northeastern University and Brigham and Women’s Hospital, focusing on representation learning for compound-protein interaction (CPI) datasets.
  • GAGET: i developed a toolkit for evaluating genome assemblies using graph representations.
  • Electric Network Analytics: I designed domain-specific metrics and web-based tools to analyze structural vulnerabilities in power grids.

đź’Ľ Industry Experience

Signal Labs

From June to August 2026, I worked as a Causal Inference Research Intern at Signal Labs in Chicago, on causal discovery outside the setting where it is usually benchmarked — real observational data, with no ground-truth graph to score against.

  • I applied constraint-based and score-based discovery algorithms to real datasets, isolating the failure modes absent from synthetic benchmarks: no ground truth to evaluate against, identifiability assumptions that are hard to justify, and structure recovery that shifts across resamples.
  • I used LLMs as a source of causal background knowledge — variable semantics, plausible relations, edge orientations — encoding it as priors and constraints to narrow the equivalence class the data alone leaves ambiguous.
  • I built an end-to-end pipeline for preprocessing, algorithm execution, and structure evaluation, enabling reproducible sweeps across datasets, algorithms, and prior configurations.