AI security conversations often focus on prompt injection, hallucinations and data leakage. Those risks are real, but attackers do not need to manipulate a model if they can compromise the infrastructure surrounding it. Microsoft recently documented compromises involving a LiteLLM gateway, a RAGFlow deployment and a Kestra workflow environment. The affected products perform different functions,…

Predictive AI becomes much easier to understand when you stop treating machine learning as a black box and start treating it like an engineering process. That is the purpose of this ML.NET exercise. In this lab, we use C#, .NET, ML.NET, AutoML, and the Ames Housing dataset to predict house sale prices. But the real…

A predictive system can fail without crashing. That is one of the most important differences between traditional application monitoring and production predictive AI. The API may still return HTTP 200. The database may still be online. The scheduled process may complete successfully. The application may still display a prediction. And the model may quietly be…

Many organizations already have most of the technology required to build useful predictive AI applications. They have years of operational data in SQL Server. They have C# developers. They have existing .NET applications. They have authentication, logging, deployment pipelines, APIs, business rules, support teams, and established enterprise architecture. What they may not have is a…

A machine learning model can be accurate, technically impressive, and completely useless in production. It can generate the right prediction and still fail to create business value because nobody knows when to run it, where to store the result, how to handle bad data, who should act on the prediction, or what happens when the…

A machine-learning model produces a prediction. Now what? Is the prediction accurate? Is it better than the current process? Can the business trust it? Does it arrive early enough to matter? And most importantly: Does it improve a real business decision? These questions are more important than simply asking whether a model achieved a high…

NOTE: We started with this example: https://learn.microsoft.com/lb-lu/%20dotnet/machine-learning/tutorials/predict-prices-with-model-builder Predictive AI becomes much easier to understand when you stop talking about algorithms in the abstract and start experimenting with real data. That is the idea behind a new hands-on ML.NET Predictive AI exercise I have published on GitHub. The exercise starts with a straightforward business problem: Given…

For decades, becoming a better software developer meant learning more. More C# syntax. More .NET APIs. More design patterns. More framework features. More Azure services. More performance techniques. More ways of solving the same problem. Those skills still matter. But Large Language Models (LLMs) such as ChatGPT, Claude, GitHub Copilot, and other AI coding tools…

When businesses begin experimenting with Predictive AI, it is easy to focus on the algorithm. Which model should we use? Should we use gradient boosting? A neural network? A specialized forecasting model? Something in Azure Machine Learning? A model trained with AutoML? Those are legitimate technical questions. But they are often not the most important…

Most businesses have spent years—or decades—collecting data. Sales transactions. Customer orders. Inventory movements. Production records. Equipment telemetry. Service calls. Project histories. Financial transactions. Staffing levels. Delivery times. That historical data describes what happened in the past. A forecasting system attempts to use those historical observations to answer a more valuable question: What is likely to…