Jean-Pierre Riehl, Innovation Director, AZEO
Although I think the word is mainly overused, the concept behind is truly a revolution changing the IT market and beyond.
AI is silently invading the source code of software, products, and service developed by many companies.
More than "magic algorithms" AI must be seen through its subtle usage that change the way we design everything. We move from a world of certainty to a world of uncertainties. And that uncertainty opens a plethora of opportunities.
The most invisible AI is what is called as "Infused AI". You can find it as of little features in your day-to-day tasks. For example, Microsoft put tons of them in all their products. A live translator in PowerPoint presentations, a threat detection in Azure technologies, automatic insights on your data with Power BI, etc. All of them benefits from massive AI investments from Redmond firm.
Behind the scenes, there are predictive functions computing data to propose to the user’s most probable actions or answers.
That leads our classical applications to become Intelligent Applications. Any label becomes and insight. Any button calls a prediction function. Any form or screen is personalized with user's data.
Then, AI is also democratizing on "things". The era of Edge Computing has begun few years ago and AI now executes everywhere, from smartphones to any connected objects. You can run AI algorithms on low-cost chipsets and any manufacturers puts dedicated AI routines in hardware (CPU, TPU, NPU, FPGA) and operating systems (WinML, CoreML, MLKit).
Changing the Way we Think and Design Software
At AZEO we inject pieces of AI in all our development processes. We don’t develop applications, we create Intelligent Apps for our customers. We don’t implement IT systems, we build intelligent solutions.
That AI revolution changes a lot of things in the software processes and we have embraced them. This is reflected by some convictions we apply systematically:
• More design thinking, more envisioning in preliminary phases of projects.
• More “AI” expertise needed to catch and distillate ideas and propose paths to implement it
• Process is definitively iterative. Sometimes, you don’t know if it is possible neither how well it can be reliable. Test and learn, fail-fast are good practices to ensure success.
• That’s why expertise is needed. Anticipate caveats, make more accurate choices, find shortcuts and lead to effective results
Concretely, we organize ideation workshops with our customers to add innovative features to their solutions. We add predictive functions in the applications to transform it into intelligent Apps.
We promote natural user experience (NUX) using chatbots, RPA, speech, augmented reality, computer vision, mobility, etc. We aim to add cognition to applications.
To make it possible, we have a dedicated data & AI practice that build predictive models, following data science processes (like TDSM). Then we industrialize and use models in apps we develop.
Different Types of AI Projects
We classify projects involving AI into 3 types.
• Call Cognitive Services
This is the easiest and fastest way to infuse AI into an application. Microsoft provides a large set of cognitive functions: OCR, computer vision, speech-to-text, text-to-speech, translation, sentiment analysis, etc. It is as easy as a simple HTTP call and it can be integrated in any application.
We inject that features behind a button click or a screen load to bring intelligence to users.
Besides these functions, there is decades of breakthrough research of Microsoft into artificial intelligence. Thanks to its clouds and services (Office 365, Xbox, Bing, Dynamics, Azure, etc.), Microsoft first to achieve human parity in many areas. That makes these features credible in the real life.
• Extend existing predictive models
But sometimes, functions provided turnkey by Microsoft are not enough. You may need to specialize some of them to fulfill your application objectives.
Microsoft proposes different ways to customize their algorithms using fields of machine learning like transfer learning or reinforcement learning. Like this, you can specialize a model to recognize objects, detect predefined faces, understand specific jargon or identify voices.
With your own data, data scientists can create dedicated algorithms that will be used as a service in applications.
For more traditional predictive models, automated machine learning (AutoML) is a powerful approach to build it. You can automate creation of algorithms directly into an application, using the power of Azure to train it. Note that Data Science specialists are required to parameterize that process.
• Create your own predictive models (from scratch)
Specialization of algorithms doesn’t cover all the scope of cognitive computing. So, building custom machine learning (ML) algorithms is still a way we need to explore. In addition to data science, data mining or statistics expertise, we need data, a lot of data. That’s why data lake projects are important for enterprises. Data engineers build pipeline to ingest and organize data and data scientists implement models against it.
They need skills and technologies. Microsoft provides a whole platform to build AI solutions.
Services
It corresponds to the first type of AI project explained above. Microsoft enriches its set of services permanently. We think that subject matter expertise is required to master any cognitive functions. That expertise differs on each cognitive field: speech, vision, language, etc.
Infrastructure and platform
Microsoft provides infrastructure and technologies to help data engineers and data scientists to build AI solutions.
It starts with hardware. Azure Cloud offers all the computable resources you need: CPU, GPU, NPU and FPGA. You have access to raw compute power, but you can access it through many platforms like HADOOP, Databricks, Containers, etc. Everything you need is available on-demand and scalably.
Tools
Microsoft is not only a platform provider for compute and execution. Microsoft provides tools to develop and operationalize data science processes.
Developers get tools to implement like Visual Studio, Azure ML Studio or Notebooks. Microsoft contributes to many frameworks, open source or house-made. We can mention ONNX initiative .Microsoft associated to Facebook, Amazon, Intel and many others to develop and support an open ecosystem for interchangeable AI models.
Then, Microsoft is making a lot of effort to bring DevOps to Machine Learning. It’s called MLOps. Even if your data science team is talented, you must industrialize model development and distribution for your apps. At AZEO, we mix our competencies on ALM (Application Lifecycle Management) to Data and AI.
Opportunities & Benefits
There are an ocean of opportunities and a lot of benefits considering AI. AI adds value to your applications, your products, your processes.
It is all about “Empowerment”: engage customers, optimize operations, empower employees, transform products.
Conclusion
With AI and data science, we drive innovation of our customers adding innovative features. We help them to build applications, solutions, IT systems that encourages new usages and products.
At AZEO, we do not consider AI as something standalone. AI serves usages, applications, users and it is only possible if you consider something broader. That’s why we infuse AI in all our competency areas.
This is the easiest and fastest way to infuse AI into an application. Microsoft provides a large set of cognitive functions: OCR, computer vision, speech-to-text, text-to-speech, translation, sentiment analysis, etc. It is as easy as a simple HTTP call and it can be integrated in any application.
We inject that features behind a button click or a screen load to bring intelligence to users.
Besides these functions, there is decades of breakthrough research of Microsoft into artificial intelligence. Thanks to its clouds and services (Office 365, Xbox, Bing, Dynamics, Azure, etc.), Microsoft first to achieve human parity in many areas. That makes these features credible in the real life.
• Extend existing predictive models
But sometimes, functions provided turnkey by Microsoft are not enough. You may need to specialize some of them to fulfill your application objectives.
Microsoft proposes different ways to customize their algorithms using fields of machine learning like transfer learning or reinforcement learning. Like this, you can specialize a model to recognize objects, detect predefined faces, understand specific jargon or identify voices.
With your own data, data scientists can create dedicated algorithms that will be used as a service in applications.
For more traditional predictive models, automated machine learning (AutoML) is a powerful approach to build it. You can automate creation of algorithms directly into an application, using the power of Azure to train it. Note that Data Science specialists are required to parameterize that process.
• Create your own predictive models (from scratch)
Specialization of algorithms doesn’t cover all the scope of cognitive computing. So, building custom machine learning (ML) algorithms is still a way we need to explore. In addition to data science, data mining or statistics expertise, we need data, a lot of data. That’s why data lake projects are important for enterprises. Data engineers build pipeline to ingest and organize data and data scientists implement models against it.
They need skills and technologies. Microsoft provides a whole platform to build AI solutions.Services
It corresponds to the first type of AI project explained above. Microsoft enriches its set of services permanently. We think that subject matter expertise is required to master any cognitive functions. That expertise differs on each cognitive field: speech, vision, language, etc.
Infrastructure and platform
Microsoft provides infrastructure and technologies to help data engineers and data scientists to build AI solutions.
It starts with hardware. Azure Cloud offers all the computable resources you need: CPU, GPU, NPU and FPGA. You have access to raw compute power, but you can access it through many platforms like HADOOP, Databricks, Containers, etc. Everything you need is available on-demand and scalably.
Tools
Microsoft is not only a platform provider for compute and execution. Microsoft provides tools to develop and operationalize data science processes.
Developers get tools to implement like Visual Studio, Azure ML Studio or Notebooks. Microsoft contributes to many frameworks, open source or house-made. We can mention ONNX initiative .Microsoft associated to Facebook, Amazon, Intel and many others to develop and support an open ecosystem for interchangeable AI models.
Then, Microsoft is making a lot of effort to bring DevOps to Machine Learning. It’s called MLOps. Even if your data science team is talented, you must industrialize model development and distribution for your apps. At AZEO, we mix our competencies on ALM (Application Lifecycle Management) to Data and AI.
Opportunities & Benefits
There are an ocean of opportunities and a lot of benefits considering AI. AI adds value to your applications, your products, your processes.
It is all about “Empowerment”: engage customers, optimize operations, empower employees, transform products.
Conclusion
With AI and data science, we drive innovation of our customers adding innovative features. We help them to build applications, solutions, IT systems that encourages new usages and products.
At AZEO, we do not consider AI as something standalone. AI serves usages, applications, users and it is only possible if you consider something broader. That’s why we infuse AI in all our competency areas.
Read Also
Choosing Microsoft Dynamics 365
Andrew Ho, CIO, Global Strategy Group
Microsoft Teams Is Growing Like Wildfire: How To Not Get Burned
Robert Mulsow, VP Technical Solution Professionals EMEA, AvePoint; Microsoft MVP
Building mHealth Apps
Neil Polwart, Group Head of Mobile, BBI Group
The Needed Change on the Software Bot Industry
Guillermo Amodeo, Head of Software Engineering, Kolokium Blockhain Technologies SL
