Bart Czernicki, Technical Architect - Advanced Analytics & AI, Azure SaaS ISV Solutions
The Microsoft Artificial Intelligence platform is a comprehensive software ecosystem that allows anyone to become a professional AI developer. At a high level, this platform consists of three key pillars: services, infrastructure, and tools. Microsoft offers these on Azure public cloud, hybrid, and on-premises deployed software environments. In this brief article, we’ll walk through the key unique value of the Microsoft AI platform: empowering organizations with varying data science, machine learning (ML), and advanced analytics skill sets to dramatically accelerate the development of AI solutions. Be sure to check out more details on the Microsoft AI Platform here.
Pillars of the AI platform
Above, you can see highlights of the exclusive functionality of the aforementioned Microsoft AI platform pillars. The first pillar is AI Services, which contains a portfolio of developer APIs. These APIs are exposed in a variety of ways that allow crafting of custom ML and data science pipelines—from pre-made, ready-to-use endpoints (i.e. Facial recognition REST API) to advanced functionality (i.e. Azure ML Services). These AI Services are offered as both software (SaaS) and platform (PaaS) Azure cloud software, abstracting the complexity of managing servers, security, compliance etc.
However, occasionally software architecture governance demands the flexibility of fine-tuning the deployments in great detail. This is where the second pillar, AI Infrastructure, comes in. It provides the core compute, storage, and networking options to build and serve all types of AI workloads. For example, Microsoft AI infrastructure includes the option to train complex neural network architectures faster, using specialized GPU hardware.
But the process doesn’t end there. An AI developer needs a professional suite of developer software to materialize and glue all of the AI ideas. This functionality is provided in the third and final pillar, AI Tools. AI Tools provides a comprehensive set of IDEs, SDKs, deep learning frameworks, and cross-platform tooling to craft AI software as they choose. This is generally the software UIs and code frameworks that AI developers will be looking at on their screens.
Empowering all AI skill sets
Now that you are introduced to the Microsoft AI Platform, let’s visualize how this can help any type of developer. The Artificial Intelligence paradigm has been around for quite some time, however it only just recently gained tremendous traction in the software domain. After all, what software team doesn’t want an automated intelligent system that can work 24/7, and doesn’t mind working the weekends? Even with the demonstrated value of AI systems, some executive decision makers are apprehensive about jumping into AI. Certain organizations feel they can’t implement AI software because they don’t have super-advanced skills, nor do they employ PhD statisticians. Conversely, mature advanced analytics organizations who have been doing statistical modeling for decades sometimes have the impression that AI conventions don’t warrant a shift from their proven practices.
As you saw earlier, the Microsoft AI platform offers a wide variety of advanced analytics functionality. Let’s look at this functionality as an “ease of AI use” and customization cross-section spectrum. The diagram below shows Microsoft AI functionality that’s easier to invest in and offers less control on the left-hand side, and functionality that requires a deeper AI investment while offering full operational control. As you can see, there are quite a lot of service and tool offerings across this cross-section.
As you look at the above diagram, note two key points: the “ease of AI use” spectrum (shown in purple) and customization options, split between “Consume” and “Build your own.” This breakdown helps align AI service offerings with the organization’s analytics comfort level.
Let’s look at some use cases using this diagram. For example, if an organization just wants to leverage production-ready models, they would ideally look at the left-hand side “Consume” tree nodes and gravitate to Cognitive Services or pre-trained CNTK/TensorFlow models. These pre-trained models can be used out-of-the-box to augment existing software very easily. However, if an organization wants to build their own models using their own data, the Microsoft AI platform provides wide spectrum of easy-to-use AI. Looking at the “Build your own” node above, teams that have basic data science skills will naturally gravitate to services like Azure Machine Learning Studio. Conversely, a team of PhD statisticians will be on the far-right side, looking for the platform to help them build complex neural network architectures.
Next steps
If you were apprehensive about getting started with AI, I hope this blog post showed how infusing your software with AI can be frictionless with the Microsoft Artificial Intelligence platform. The Microsoft AI platform provides a wide variety of services, infrastructure, and tools that help all kinds of information worker personas. Whether you are a graduate student, data steward, application developer, or a professional data scientist, the Microsoft AI platform has a functionality for you!
In part 2 of this article, we’ll explore how this wide complexity of AI services can be allocated to different AI information worker personas with multiple detailed examples.
Register for the Data & AI Community call on May 4 at 10 am PT
Data & AI Partner Community
from TechNet Blogs https://ift.tt/2Fy43wm
No comments: