Twenty-five years of making things work, now pointed at AI.
I started as an electrician, spent six years keeping wind turbines turning, and came to artificial intelligence through a degree taken later than most people take it. That order matters: I learned how physical systems fail before I learned how to build software ones, and it shows in how I build.
- Based in
- Sardinia, Italy
- Research
- 2 peer-reviewed papers
- Degrees
- BSc & MSc, both 110/110 cum laude
- Hands-on since
- 2001
What I do now
I work across the full AI stack — data analysis, machine learning, deep learning and large language models, from training and fine-tuning through inference, evaluation and deployment. In practice that means I am as comfortable writing the evaluation harness as the model code, and more interested in whether a system holds up in production than in whether it looked good in a notebook.
My day job is municipal IT: network, infrastructure and security for a public administration. It is not glamorous, and it is the single best thing that ever happened to my engineering judgment. Running infrastructure that other people depend on teaches you what "it works on my machine" is actually worth.
Alongside that I build and publish — multi-agent LLM pipelines, DeFi security tooling, XR interaction work, fine-tuning studies — and write them up properly, because a result nobody can reproduce is not a result.
Experience
-
Oct 2025 — Present
Information Technology Officer
Public administration · Sardinia, Italy · Full-timeI run the technical infrastructure behind a municipal administration serving roughly 130 civil servants and the community around them — network, systems, security and compliance, alongside the slower work of modernizing public-sector IT so that services actually get better rather than just newer.
IT operationsNetwork securitySystems administrationPublic sector -
Sep 2024 — Sep 2025
Research Fellow
Università degli Studi di Cagliari · Cagliari, Sardinia · Part-timeApplied research on computer vision and large language models. Two peer-reviewed papers came out of this period: a roadside monocular 3D object detection framework, and a system that turns natural-language questions from students, instructors and managers into knowledge-graph queries.
Computer visionLarge language modelsKnowledge graphsBenchmarking -
Feb 2024 — Sep 2024
Data Analyst
The Cloud Alchemist · Cagliari, Sardinia · Part-timeBig-data analysis and applied AI for cost prediction and cost optimization, plus coordination of the technical work behind it.
SQLBig dataForecastingCost optimization -
Sep 2021 — Sep 2024
Data Analyst
Independent · Cagliari, SardiniaAnalysis and modelling work taken on directly alongside my degree — the period where the statistics stopped being coursework and started having consequences.
SQLPythonLarge language modelsReporting -
Aug 2004 — Aug 2010
Wind Turbine & Blade Repair Technician
IVPC Eolica · Bortigiadas (SS), ItalyMaintenance, fault-finding and repair on V47 and V52 turbines — electrical, electronic, hydraulic and mechanical — and team leader for the blade repair crew, working both on the ground and at height. Over 100 machines a year in routine maintenance, 95% of faults cleared inside 24 hours, and better than 99% availability across a 35-unit fleet.
Fault diagnosisTeam leadElectrical & hydraulic systemsField maintenance -
2001 — 2003
Electrician
Ditta Piras · Ditta Uleri · Sassari and Ploaghe, ItalyDesign, installation and maintenance of civil and industrial electrical systems, public lighting, and refrigeration and heat-pump plant — plus the troubleshooting that comes with all of it.
Industrial systemsInstallationTroubleshooting
Education
Master's Degree, Artificial Intelligence
Applied Artificial Intelligence within the Master's in Computer Science. Thesis: "Agentic Planning for Extended Reality Training Scene Generation from Structured Natural Language Specifications", supervised by Dr. Mirko Marras — multi-agent LLM planning for XR industrial training scenarios, and the benchmarking methodology to tell whether it worked. Coursework: Deep Learning, NLP & Knowledge Graphs, Reinforcement Learning, Information Retrieval, Data Mining, Advanced Data Management.
Bachelor's Degree, Artificial Intelligence
Applied Computer Science and Data Analytics. Spokesperson for the student body for two years, representing students in university meetings. Foundations in convolutional and recurrent networks, data analysis, and software engineering.
Electrotechnical and Automation
Expert electrician for electrical engineering and automation — calculation, design and development of industrial and domestic electrical systems, and the automation layered on top of them.
Research
2 peer-reviewed papers, both from the research fellowship at the University of Cagliari.
WAILS 2025Facilitating Information Extraction in Education by Translating Questions from Students, Instructors, and Managers Into Knowledge Graph Queries Through Large Language Models
doi:10.1007/978-3-032-17604-2_12 ↗ ACM UMAP 2024 · Adjunct ProceedingsRoadSense3D: A Framework for Roadside Monocular 3D Object Detection
doi:10.1145/3631700.3665236 ↗Get in touch
Open to AI/ML work, collaborations, or a good technical conversation.
The most reliable way to reach me. I read everything and answer properly.
marcosau.dev@gmail.comFull professional history, roles and education, kept current.
/in/marcosauGitHub
Source for the projects on this site, plus smaller experiments.
@Marco-SauYouTube
Technical walkthroughs on AI, data engineering and the tooling around them.
@MarcoSau1981