Pietro Mascolo

ML systems rarely fail dramatically: they fail quietly. This session cuts through the noise and shows the small set of monitoring signals that prevent the majority of real-world model failures. We’ll build a practical monitoring architecture in Python using Evidently, River, FastAPI, and MLflow,...

Monitoring is overhyped and badly implemented: I’m going to show you the five signals that prevent most ML disasters, cand how to implement them in plain Python. You’ll leave knowing exactly what to monitor, what to ignore, and how to build a reliable stack.

Graphs are a powerful way to represent and analyze relationships in data, from social networks to molecular structures. This talk delves into advanced graph-based machine learning techniques using Python libraries like NetworkX, PyTorch Geometric (PyG), and StellarGraph. We’ll cover the...

Sequence modeling plays a crucial role in machine learning applications, from time series forecasting to natural language processing and healthcare analytics. This talk introduces RETAIN (Reversed Time Attention Model), a powerful architecture designed for interpretable sequence modeling in...

Graphs unlock insights from complex relationships in data. Learn how to use Python libraries like NetworkX and PyTorch Geometric for advanced graph-based machine learning applications, from fraud detection to recommendation systems and beyond.

Working in Data Science is already not simple when we are in controlled conditions with a fixed training and test dataset, but how can we cope with the real world? What if we have no control over the training data? What if there is no defined test set? What if we do not have ground truth? What if...

We will show how to extract information and details from ML/AI models, using SHAP and embeddings, in order to understand how they learn and how we can leverage this to ensure robustness and evaluate models before they are deployed and used.