ai and machine learning

I share my mixed experiences with ChatGPT, noting that while it excels at specific technical tasks like data conversion, it often provides unhelpful advice for broader questions. To help you build more robust solutions, I also introduce Azure OpenAI as a powerful commercial alternative with enterprise capabilities and guide you through the initial steps of provisioning it in the Azure Portal.

I explain how to leverage Azure OpenAI’s PaaS service to train GPT models on your own custom data by creating simple JSONL dataset files. This approach allows you to move beyond generic responses and tailor the AI’s behavior for specific, high-value use cases.

If you are curious about how accessible modern machine learning has become, this post shares my hands-on experience building an image classification model using Microsoft’s visual tool, Lobe. I walk through the process of labeling and training a model to distinguish gym weight plates, highlighting both the software’s intuitive interface and the practical challenges I encountered when dealing with ambiguous data and unexpected crashes.

Discover how the latest Lobe update simplifies deploying machine learning models directly into Power Platform by eliminating the need for complex TensorFlow integration. I walk you through creating a fruit detection model using Microsoft’s sample images, labeling them in Lobe, and preparing them for seamless publication to AI Builder.

If you are looking to proactively detect cost anomalies in your Azure environment, this post highlights the new Cost Insight feature that helps prevent unexpected spending spikes. I walk you through how to enable this free, machine learning-driven preview tool in the Azure Portal and demonstrate how it identifies deviations in your billing data over the last 60 days.