Andrea Maurino

Over the course of my career as a Full Professor of Computer Science and technology advisor, I have coordinated development and technology innovation teams across academic, corporate, and public sector engagements, combining strategic advisory with hands-on technical leadership.

My role has consistently spanned the full innovation lifecycle: defining data and AI strategies aligned with organizational goals, designing system architectures, and overseeing delivery to ensure solutions were implemented on time, within scope, and to a high technical standard. I have led multidisciplinary teams spanning data engineering, machine learning, and software development, coordinating work across research groups, industrial partners, and public administrations.

Representative engagements include technical coordination of an AI-driven credit risk platform built on PSD2 open banking data for a national credit consortium; architecture and strategy definition for open data platforms and NoSQL infrastructure for regional government bodies; and design oversight of AI-based document intelligence pipelines for Italy's national statistics institute. I have also co-founded two technology ventures, translating research outcomes into deployable products and guiding their technical direction from concept to market. This advisory and delivery work is underpinned by an active research and teaching practice: I lead funded research programs (having managed approximately €3.5M in project funding), publish extensively on data quality, data-centric AI, and generative AI applications in finance (120+ peer-reviewed publications, H-index 27), and teach data management and AI courses at graduate level. This dual grounding in rigorous research and applied delivery allows me to bridge strategic vision with technical execution — assessing emerging technologies critically, translating them into concrete architectures, and ensuring teams deliver reliably against ambitious goals.

Alongside my academic and advisory work, I have hands-on experience as a Python developer, designing and building open-source libraries and orchestration tools to support rigorous, reproducible machine learning experimentation.

I designed and published PuckTrick, a Python library (currently v1.0.4, distributed via PyPI) for deterministic, controlled data corruption. The library enables systematic injection of realistic data quality issues, including label noise (NCAR/NAR/NNAR), outliers, missing values, generic noise, duplicates, and data drift — into datasets, supporting robustness testing of machine learning models under controlled degradation conditions. I implemented multiple accumulation modes (composed, extended, new) to model how errors compound over time, multiclass label perturbation via configurable flip matrices, and a modular drift-injection component supporting seven distinct drift operators.

Building on this, I developed Dirtify, an orchestration layer that runs large-scale experiments combining PuckTrick's error-injection logic with multiple datasets, models, and error types. Dirtify was engineered to run on a Slurm-managed HPC cluster, executing tens of thousands of model training runs (50,000+) reliably at scale. I implemented DuckDB and CSV-based checkpointing for fault tolerance and resumability, and resolved complex dependency chains across Python 3.11, PyCaret 3.3.2, scikit-learn 1.3.2, and hdbscan, ensuring stable execution across a shared, resource-constrained research computing environment.

This development work required end-to-end engineering ownership: package design and API usability, performance optimization for large-scale batch execution, dependency and environment management, and debugging of distributed, long-running jobs, skills directly transferable to production data engineering and MLOps contexts.

Alongside my academic and technical work, I have extensive experience designing and delivering training programs on data management, data quality, and AI, both within university master's programs and directly for companies and public administrations.

As a Full Professor at the University of Milano-Bicocca, I have taught Data Management and Data Visualization courses across multiple master's-level programs for over a decade, and served for six years (2019–2025) as President of the Board of the Master's Degree Program in Data Science, where I contributed to curriculum design across data warehousing, data lakes, pipeline patterns, data quality and data contracts, data-centric AI, and RAG/agentic AI architectures.

Beyond academia, I have delivered corporate training engagements for a range of organizations, including Engineering S.p.A., BIP, Lazio Crea, Umbria Digitale, Datlas, CTC (Consorzio Tutela del Credito), and Sole 24 Ore. These sessions have covered practical topics such as data quality methodologies, open data and NoSQL technologies, and the application of AI and generative AI in operational and regulatory contexts — always tailored to the specific technical maturity and business needs of the audience, from technical teams to non-technical management.

My approach to training combines rigorous theoretical grounding with concrete, hands-on examples drawn from real research and industry projects, including tools I have personally built (such as PuckTrick and Dirtify) and consulting engagements I have led (such as AI-driven credit risk systems and open data platforms). This allows me to make complex technical topics accessible and immediately actionable for participants, whether the goal is upskilling a technical team or building strategic literacy among decision-makers.