I'm looking into the possibility of using Azure Machine Learning (Azure ML) workspace for implementing AI agents. My goal is to showcase a complete AI agent workflow using only Azure services without depending on any open-source platforms. I'm particularly interested in developing and orchestrating these agents programmatically, steering clear of low-code solutions or user interfaces. I want to know if Azure ML is the right environment for this or if I should consider using a traditional IDE like VS Code or PyCharm instead. Any insights would be appreciated!
2 Answers
Azure ML is mainly for training models. If your focus is purely on creating an AI agent, then Azure AI Foundry is where you should be looking. If you just want to work with an OpenAI model from your code, consider using Azure OpenAI. You can deploy models and build an application around it using an API key from your OpenAI deployment.
It sounds like you're better off looking into Azure AI Foundry instead of just Azure ML. The ML workspace is mainly for training models, not really for building agents. If you're specifically interested in agentic AI, check out the Microsoft Agent Framework as well. That might give you what you need!
Yeah, I was hoping to run my code on the ML workspace and connect it to AI Foundry for the Azure AI services.
I think he might be interested in building something like a transformer model too!