Last week was the annual Microsoft Build conference in Seattle. Build is the main global Microsoft developer conference and this year many sessions covered Microsoft’s progress in the areas of AI and ML.
I spent some time browsing the online session videos and was overwhelmed by the volume of content from the 3 day conference. I therefore decided to curate the AI/ML sessions and create a blog post (mainly for my own reference). Enjoy!
Microsoft AI Platform: State of the Union
This session gives a good overview of the current state of the various AI/ML services offered by Microsoft and is a good place to start:
What’s new with Azure Machine Learning
The latest news on what’s new since the last set of AI/ML updates in September:
Microsoft AI overview for developers
This is a great overview, especially the demo by Seth Juarez. One of the tools mentioned is Netron, a visualizer for deep learning and machine learning models, definitely one to check out!
Demystifying Machine and Deep Learning for Developers
This session is a high level overview of ML for developers.
Machine Learning at Scale
AI/ML engineer Paige Bailey covers Azure DataBricks, Batch AI, KubeFlow + AKS, Stream Analytics and Event Hub:
Deep Learning at Scale
Scaling deep learning with Azure, showing how you can use frameworks like Tensorflow, MXNet, PyTorch, and Caffe, and take advantage of elastic GPU enabled hardware (on Azure).
This session included a couple of interesting announcements:
- New GPU instance coming soon, which will be the most powerful GPU instance on Azure, the ND v2. This will host 8 NVIDIA V100 GPUs with 16GB GPU memory, Skylake CPU architecture and NVLink interconnect.
- Some new Azure Batch AI APIs including a capability to group jobs into experiments and then group experiments into a workspace. A workspace looks like it will be a persistent container and will be useful to group together dev/test/live experiments etc.
Building Custom AI Models on Azure using TensorFlow and Keras
Using ML.NET, .NET developers add custom AI into existing .NET apps.
Fun ways to Explore ML.NET
A short session on how to use ML.NET to bring custom machine learning to your .NET applications.
Migrating Existing Open Source Machine Learning to Azure
Learn how data scientists can transition their existing workflows — while using mostly the same tools and processes — to train and deploy machine learning models based on open source frameworks to Azure.
Get Productive with Python Developer Tools
Using AML Python SDK
The Azure ML SDK for Python provides a single control plane API to the data scientist to execute the key AML workflows of Provisioning Compute, Model Training, Model Deployment and Scoring entirely in Python.
Getting Started with Visual Studio Tools for AI
Visual Studio Tools for AI makes it easy to train, debug and deploy AI applications and services.
AI Platform Discussion
A 30 minute round table discussion on the state of AI development, together with a QA session: