Join the AWS Dev Day Belgium on November 13 in Mechelen
Enterprises are looking for efficiency and the desire to gain and keep a competitive advantage through faster innovation using Artificial Intelligence technology. Yet, there aren’t enough IT professionals with the necessary expertise in the cloud.
One of Amazon Web Services’ (AWS) mission is to make machine learning accessible to all developers by removing the barriers to build these systems. It is now possible to integrate machine learning services directly into your mobile- and web-application by calling application programming interfaces (APIs) for AI services like Amazon Rekognition, Amazon Lex, and Amazon Polly. All are intended to put the power of computer vision, conversational agents, and text to speech into the hands of all developers. That increases the accessibility: now, how can you ensure that machine learning accelerates your time to market?
Patrick Sard, Solutions Architect at Amazon Web Services (AWS), has many years of experience with cloud and cloud services. “The cloud makes it easier to quickly build, deploy, and scale your machine learning (ML) models. Additionally, our AI services help you incorporate image recognition, video recognition, natural language processing, and speech to text in your app without requiring deep knowledge of AI or ML. AWS has the broadest range of AI services and has released more than 200 machine learning features and capabilities in 2018 alone. In fact, 85% of TensorFlow projects in the cloud run on AWS,” he says.
Building your Machine Learning capability
The AWS machine learning stack consists of three layers: top, middle, and bottom.
• The top layer has ready-to-use AI-services based on pre-built data models. Ideal if you miss deep ML expertise. There are six categories of services: vision, speech, language, chatbots, forecasting, and recommendations. Each of these categories has one or more services (see figure 1). You bring along your data and effortlessly fuse these services into your application. In other words, you consume the AI services immediately without having to think: how do I build a model first?
• The middle layer of services, Amazon SageMaker, focuses on data scientists and developers with prior knowledge in ML. You conveniently build, train, and deploy better machine learning models at scale. How? By using pre-installed optimized algorithms (or bringing your algorithm if you prefer), one-click training, and automatic model tuning to achieve the most accurate predictions. Last but not least, one-click deployment ensures that deploying your model becomes a straightforward task.
• The bottom layer of the stack handles the complete infrastructure such as servers, compute power, and storage. It is also the layer where you manage the installation and deployment of machine learning frameworks. Patrick: “This is for engineers who want to set up the infrastructure and machine learning framework themselves.”
Figure 1: The three layers of the AWS machine learning stack
To better understand the advantages of machine learning for your business, Patrick shares three examples of customers who are already using it in their applications.
SkinVision, a Dutch company, built a mobile app to self-assess your skin health. One out of five Belgians is having skin cancer at some point in their life. Therefore, it is crucial to monitor your skin. However, that is easier said than done. The SkinVision self-assess app allows you to take pictures of your skin, and the AWS machine learning-based algorithms assess those in thirty seconds. After that, staff dermatologists examine the image to control accuracy.
Patrick explains: “Although the app does not make an immediate diagnostic, it indicates if your skin cancer probability is high, medium, or low. That helps you to determine if you have to schedule an appointment with the dermatologist. Besides, the data collected by the app provides dermatologists better insights into skin cancer. The app even has the option that a dermatologist contacts you if your skin cancer probability is high or medium. SkinVision has already analyzed more than a million different images and uses this to improve the accuracy of its data model. The Department of Health and Social Care in the United Kingdom recognizes the SkinVision app as a means to self-assess the probability of skin cancer.”
Convoy, a company in the United States, realized that 40% of the trucks on the road are empty (by the way, the same proportion also applies to Europe). “To improve the truck’s efficiency, they came up with the idea to create a platform that better matches shippers and drivers. So, they created a mobile app, a digital online marketplace that uses AWS AI services to analyze shipping jobs along with trucker availability. That gives them the information to recommend matches that are timely and cost-efficient. As Convoy works with many shippers and drivers, its AI can take advantage of more data from the entire freight network for demand forecasting,” says Patrick.
Davinci, a financial organization, based in the Netherlands, has a software as a service for loan and mortgages. They provide this software to financial institutions in the Netherlands. The customers of these institutions use it for the online processing of their loans and mortgages requests. Patrick: “If you apply for a mortgage, you have to upload official documents like your identity card and salary fiche to support your application. Davinci infers the data on these documents with the AWS machine learning services. The algorithm learns how to spot the right information in the right place and optimizes the data model accuracy. That helps to reduce the repetitive manual tasks performed by the financial professionals, although they are still in charge of the final decision to approve a loan or mortgage”.
Do you want to accelerate your time to market by putting machine learning in the hands of every developer? Register for the AWS Dev Day Belgium on November 13 in Mechelen.