Galois
A database system that uses LLMs as a storage layer for SQL queries, with custom physical operators and query optimizations. Published at SIGMOD 2025.
Hi, I'm Sara 🌸
I'm a first-year PhD candidate in machine learning for healthcare in Prof. Björn Eskofier's group at LMU Munich, funded by the Munich Center for Machine Learning (MCML).
Originally from Italy, I studied Computer Engineering and Data Engineering at Politecnico di Torino and Intelligent Communication Systems at EURECOM, as a double degree. Before the PhD, I spent 18 months at SLB in Cambridge, MA, building ML pipelines for methane emissions monitoring.
Most of my work lives where physics-based models meet neural networks: taking noisy real-world sensor signals and making them trustworthy, first for the environment and now for health.
News
- Started my PhD in Björn Eskofier's group at LMU Munich, funded by MCML 🇩🇪
- Wrapped up 18 months at SLB on methane emissions monitoring and moved from Boston to Munich.
- Logical and Physical Optimizations for SQL Query Execution over Large Language Models published at SIGMOD 2025 in Berlin.
- Joined SLB in Cambridge, MA as a Data Engineer 🇺🇸
- Graduated from the double MSc at Politecnico di Torino and EURECOM, with a thesis on deep learning anomaly detection done at Qualcomm.
- Started an internship at Qualcomm in Munich.
- Moved to Sophia Antipolis for EURECOM with an Erasmus+ scholarship 🇫🇷
Career
Experience
-
PhD Candidate, LMU Munich & MCML
Machine learning for healthcare with wearable and multimodal sensor data, in Björn Eskofier's group.
-
Data Engineer (VIE), SLB
Built a methane detection pipeline end to end: a temporal convolutional network that learns a correction on top of a physics ODE for humidity inside gas sensor housings, trained with a self-supervised, physics-guided loss.
-
Data Engineer Intern, Qualcomm
Deep learning anomaly detection for semiconductor manufacturing (90% accuracy) and pipeline optimization for large-scale sensor systems.
-
Data Analyst, Politecnico di Torino
Anomaly and long-term trend detection across more than 100 sensor streams.
Education
-
PhD, Machine Learning for Healthcare
-
MSc Intelligent Communication Systems (double degree)
Erasmus+ scholarship, selected among 300 candidates.
-
MSc Data Science and Engineering (double degree)
Thesis: Enhancing Anomaly Detection in Semiconductor Manufacturing Using Deep Learning, with Qualcomm.
-
BSc Computer Engineering
Projects
ML for Communication
Replication of decentralized learning over wireless networks with broadcast-based subgraph sampling.
Semantic segmentation for autonomous driving
A federated learning benchmark for autonomous driving, plus a Clustered Averaged Batch Normalization.
MALIS
Implementations and studies of machine learning methods for intelligent systems.
📈 Numerical optimization
Constrained and unconstrained optimization techniques for large-scale problems.
Publications
Logical and Physical Optimizations for SQL Query Execution over Large Language Models
SIGMOD 2025 · Proc. ACM Manag. Data 3(3)Beyond research
-
Lead Mentor, WeAreHere · Politecnico di Torino
Mentored 60+ women in STEM bachelor's programmes and coordinated a team of 30 mentors.
-
Teaching
Teaching assistant for Linear Algebra at Politecnico di Torino, math tutor at GoStudent with video tutorials used by 200+ students, and outreach to high-school girls through the "Ingegneria Informatica è creatività" programme.
-
Community
Reply Ambassador (organized the Reply Challenge at EURECOM for 50 students), QWomen volunteer at Qualcomm, and events organizer with BEST Torino.