Machine learning models need updating - what's the reliable way to do it? While in Romania, Richard sat down with Annie Talvasto to talk about her work helping to build DevOps practices around machine learning: Building repeatable processes for data ingestions, cleaning, organization, model building, and deployment. The challenges are the arrays of skilled people needed to operate and evaluate the pipeline - it takes domain experts to know if the machine learning results are accurate and valuable. Tooling can help, but it is only in the early days. If your organization is keen to get machine learning into the company, you need to do some careful planning!

Links

• Jupyter Notebook
• Pandas
• MLFlow
• Azure Machine Learning

Recorded April 20, 2024

How's your identity governance? Richard talks to Jef Kazimer from Microsoft about Entra's capabilities to help you have robust governance around identity. Jef talks about the lifecycle of identity - when someone joins the organization and a new identity gets created, privileges changing as roles evolve, to the eventual offboarding when that person departs. The same sort of cycle exists for devices, too - the question is only how much work you must do to get through those cycles. Entra offers tooling to get identity changes out of your support tickets - make your life easier with effective identity governance!

Links

• Microsoft Entra
• Conditional Access
• Microsoft Entra Permission Management
• Road to the Cloud
• Entitlement Management
• My Access Portal
• Just Enough Administration
• Privileged Access Workstation
• Entra Lifecycle Workflows
• Azure Logic Apps
• Usage and Insight Reports in Entra
• 425 Show

Recorded April 26, 2024

How does artificial intelligence fit in with SQL Server? At the Microsoft Fabric Conference in Las Vegas, Richard sat down with Muazma Fahid and Bob Ward to discuss the AI developments in SQL Server. Muazma talks about SQL Server as a crucial source of data for building machine learning models and the new features added to make SQL Server a key store for vector data and other elements of machine learning. There's also Copilot for Azure SQL Data to help with diagnostics in your databases and to use natural language to write queries - it SQL Server Natural Language Query all over again, but certainly better than ever.

Links

• Responsible AI Practices
• Phi-2
• Retrieval Augmented Generation
• Copilot for Azure SQL Data
• JSON in SQL Server
• RegEx in SQL Server
• sp_invoke_external_rest_endpoint
• SQL AI

Recorded March 27, 2024