Italy
Passionate about Machine Learning & Data Science applied to Process Engineering.
Trusted technical advisor to Seeq customers, helping them solve complex process challenges and maximize the value from the analytics platform.
- Developed a catalyst production batch monitoring system with Aveva PI, pushing live Golden Batch trendlines back to the DCS control room to enable real-time performance comparison and early deviation detection. - Created a hybrid model based soft-sensor to monitor the concentration of a reactant fed into a polymerization reactor, ensuring consistent and reliable production of a high-performance plastic. - Developed a predictive maintenance model for electric motors, estimating winding temperatures to identify wear and fouling through overheating patterns. - Built an ANN model to predict site energy consumption from ambient conditions, operational parameters and temporal patterns, supporting utilities optimization and data-driven decision making. - Built a dashboard to automate the collection, cleaning, and visualization of utilities consumption data for a petrochemical site, providing actionable insights with minimal manual data entry.
Deployed a random forest classifier-based ML model to predict and prevent compressor failures, resulting in a reduction of GHG emissions by avoiding up to 8 hours of flaring, saving costs associated with unwasted feeds and recovering thermal energy for steam production
- Optimized the profitability of Saudi Aramco’s nationwide oil and gas plant network using hybrid modeling and real-time MINLP optimization. Conducted extensive plant data collection, analysis, and cleaning. - Developed and validated highly accurate first-principles models, then combined them with computationally inexpensive reduced-order models trained on synthetic data via PLS/Lasso regressions. - Delivered a technical seminar on surrogate modeling with gPROMS, produced supporting documentation, and supported clients in model failure troubleshooting and diagnostics.
Within the EU-funded CoPro project aimed at enhancing energy and resource efficiency in industrial plants: - Developed a steady-state detection algorithm in Python leveraging PCA. - Generated surrogate models for the KPIs of the cumene process by combining gPROMS first-principles models with Partial Least Squares Regression.
Developed a mathematical model of an industrial rotary kiln reactor for hydrofluoric acid production, performing sensitivity analysis and optimization to meet increased production targets.