Hi, my name is Barnum Bora and I am the head of artificial intelligence and machine learning for the asia-pacific region. Here at Amazon Web Services, we are very thrilled to have. You join us today for the AWS innovate online conference, machine learning special edition 2019. Now this particular session building business outcomes with AI and machine learning is intended as an introduction to how our customers are building intelligent business systems powered by AI and ml, and how these
Services, in conjunction with the large number of complementary AWS technologies, provide a great platform for our customers to build their own AI and ML powered solutions to drive business value towards the latter part of this session. We will also hear directly from one of our valued customers about their machine learning journey on AWS in our industry. A very familiar and commonly used statement is that AI is the new normal and Amazon weve been making huge investments in machine learning since 1995. Many of the capabilities, our
Customers experience within amazon com and AWS are driven by AI, so for us, AI has always been normal, but this statement also holds quantitative, meaning on a much broader context. Lets dive deep to understand why this statement holds true and why organisations, both big and small public and private are rapidly accelerating their investments into AI in machine learning by multiple estimates, the gross domestic product of the world in 2017 is known to have been approximately 75 Trillion dollars now this is a measure
Of the overall economic productivity of the entire human race, many social and analytical organizations have made growth projections for the GDP to increase rapidly over the next decade and beyond. Technological innovation has been a key enabler in the ever accelerated increase of human productivity in recent history and technological innovation stemming from ail ml, is starting to take center stage across industries and across geographical boundaries. A PwC study recently estimated that global GDP will grow rapidly to approximately a hundred
And twelve trillion dollars by 2030. This is clearly a great testament to the innovative efforts of our generation, but the same report estimates an even more exciting statistic that is of significant importance to our discussion in this session today. It is projected that about fifteen point, seven trillion dollars is going to be added to the global economy in 2030 by the productivity, improvements and net new value generated by AI and machine learning. In other words, 14 % of total human productivity by 2030 is going to
Come from AI, our customers realize the significant improvement and empowerment that AI can provide to them and their customers. Ai technologies, like computer vision, is helping doctors detect and prevent harmful diseases. Much earlier and more accurately than previously possible. Real-Time transcription and translation is helping bridge cultural and language barriers and make the world more accessible at AWS. We are striving
To empower the builders who are embracing AI now, our mission at AWS is to put AI and machine learning in the hands of every developer. Democratizing AI to all of our customers is a core mission for us, and we are seeing that just like how AI n ml has helped Amazon build a better business. Our customers are benefitting rapidly from adopting AI quickly. We are actively
Working to make sure our platform is accessible and helpful, not just to the seasoned ml practitioners and data scientists, but is also highly relevant and useful for emerging AI practitioners and developers. Now we have been using machine learning to power our business for over two decades. Many of the systems and evasions that our customers now rely on and experience everyday are powered by ml. The amazon com recommendations, engine which has been a core part of our retail business for many years, is driven by machine learning. Our Amazon, robotics division uses machine learning and
Autonomous robots to determine optimal picking routes within our fulfillment centers, in fact, our entire supply chain forecasting and capacity planning are informed by machine learning algorithms. Now, beyond those use cases ml enriches our customer experience in new and exciting areas of innovation, such as our drone delivery initiative. We call prime air or our voice based products such as Amazon, echo, which leverage Amazon Alexa services and the computer vision and sensor technologies power. A revolutionary retail experience in amazon co stores where customers can just walk out of a store without having
To wait in a checkout line now all of these efforts would not have been possible without machine learning and deep learning. In fact, today, we have thousands of Engineers at Amazon, committed to machine learning and deep learning and its always been a big part of our heritage, but driving business value with ml and the innovation with ml.
Is an iterative process every day we apply new AI machine learning, based improvements to the Amazon business at a global scale through AWS. Now one such recent improved is the edition of Amazon scout Amazon Scout is a completely new take on the recommendations. Engine recommendations typically are based on data about past purchases, user behavior, similar transactions, correlations, etc, but some of our customers want a heavily visual buying experience where they arrive at their desired items through a
Visual pathway of preferences, Amazon Scout is an engine that uses computer vision, algorithms to recommend products based on visual similarity and preferences. We encourage you to go. Try it out the same computer vision, technologies and innovations that are used to build experiences like Amazon Scout and Amazon. Go check out, less stores are available for our customers to utilize through the AWS platform, now working closely with our customers and learning iteratively from our own experience. Over the years we found that there are some chronic challenges that have historically prevented
Organizations from adopting AI and machine learning quickly, we have consistently observed that ml expertise has been historically rare, building and scaling. These ml systems has been hard. Deploying and operating models has been a time-consuming exercise and has typically been cost prohibitive. Hence, while building our services, our engineers actively try to address these roadblocks and mitigate them. The overall platform needs to be cost effective, easy to use and scalable in order to truly help our customers, customers have also
Consistently shared with us that even for seasoned practitioners and data scientists, tasks like data acquisition, labeling framework selection, scale, our training model, tuning testing simulation and inferencing at scale has always been a challenging list of tasks. Another consistent requirement both for our engineers and our customers, is that an ml solution would require training and inferencing, both in the cloud and on the edge and the experience needs to be seamless and manageable. This is one of the core capabilities we have engineered into the platform, especially with
Services like Amazon sage maker, taking this feedback and learnings as design inputs and iterating consistently. The AWS AI nml platform is designed to be accessible to every developer and the services are designed to simplify and mitigate the key roadblocks and provide developers with the capabilities that helped them build the intelligent systems that can empower their organizations, the AWS, AI and machine
Learning stack is designed with three layers: AI services, machine learning, services and ml frameworks and infrastructure. The layers correspond to the different levels of customization and configurability that different business solutions may need the top layer comprised of our AI services, which are plug-and-play API services that offer instant ml based outputs in the areas of vision, speech, language chat, BOTS forecasting and recommendations developers Can call these apis directly and integrate the AI functionality they provide directly into their solutions.
Without the need to explicitly learn and become data scientists or AI practitioners, these AI services are used by tens of thousands of our customers daily for a broad range of a I backed applications, especially the ones that require human interaction or repeatable. Similar acts like recommendations now the second layer contains the plethora of ml services and functionality that is provided to ml practitioners through our Amazon sage maker service, for situations where our customers want to build fully customized ml models and deploy them at scale. Amazon sage maker is the fully
Managed ml platform that provides the complete end-to-end ml functionality from building to training, to testing and then deployment and management. The sage maker platform greatly simplifies and accelerates the process of building ml powered systems from scratch by providing a large amount of inbuilt functionality and manageability across all major stages of machine learning development now for solutions that require full configurability of the infrastructure where customers need to
Utilize many other open source technologies. The third layer provides the most comprehensive list of compute targets and infrastructure options, both in the cloud and the edge to truly build end-to-end ml based solutions. Our customers have the option of utilizing CPUs, GPUs, FPGAs containers, serverless, architectures, elastic inferencing or even custom design, inference hardware like AWS influenza and even edge devices through AWS, IOT and Greengrass. Now it is important to remember, though, that machine learning is only a component of
The overall business solution and the depth and breadth of the broader AWS services provide the true end-to-end capabilities that an organization needs to be successful in their ml journeys. Aws provides a vast and comprehensive platform that our customers know and love, and the entirety of the platform greatly extends the capabilities of the ML solutions. Customers want to build now a common pattern. We observe across our customers and our internal business teams in Amazon is that driving business value with a line. Machine learning is a cyclic
Process that involves getting relevant data and applying domain knowledge and expertise to that data to build intelligent business systems and experiments that provide business value. This business value typically manifests itself in five broad categories of outcomes, typically systems and processes that involve a lot of repeated decisions of similar type like fraud, detection systems, etc are being automated by our customers. Using machine learning, machine learning based signals like in the case of weather forecasting systems are used by professionals to make critical decisions like planning airline routes.
Now forecasting is a perfect example of a type of business task where predictive ml is helping in looking at future behavior. Our customers are prescribing major process changes as well from ml based intelligence and simulation, that is helping them greatly improve their productivity now. Finally, we have also observed that a lot of customers identify net new capabilities to further their businesses. For example, our customers are utilizing technologies like computer vision in
Self-Driving cars to disrupt their industries, our customers are actively building innovative ml based intelligence systems to empower their businesses at a global scale through AWS. Now let us look at a few of them to see what they have been able to achieve with the platform into it. The builders of the popular accounting software QuickBooks have been able to accelerate their ml development, utilizing Amazons sage maker, resulting in a 90 % reduction in deployment time time to innovation.
And value is a game-changing capability for most organizations, and customers, like Intuit, have taken full advantage of the acceleration that the platform provides. Our customer to simple are a self-driving technology. Startup company based out of China, who have utilized the AWS ML capabilities to build a revolutionary self-driving platform that can plan the driving route, a thousand meters ahead, while on the road
Now, using Amazon sage maker Formula Ones, data scientists are training deep learning models with more than 65 years of historical race data to drive insights that they never had access to. Before now, bgl corporate the builders of the popular fund management software, simple fund 360, have built a fully AI based fund management assistant that utilizes deep learning our customer Sunday insurance, who are an inshore tech startup in Thailand, have built a fully a I based insurance underwriting Engine and are disrupting the insurance pace.
Across the region, car sales, calm, Australias number one car marketplace: utilize, the AWS AI nml platform to do automated image, classification and multiple other ML back systems to build a better experience for their customers. Now tens of thousands of companies are well and truly on their journeys. Utilizing the AWS, AI and evelle platform, the completeness of the platform and the flexibility and reliability of AWS are
Fundamental to why our customers are choosing to build their ml solutions on AWS now lets hear from one of our valued customers car sales about their ML journey with AWS. I am a Christmas nellwyn at the head of AI at Carsons. Our teams role is to research and build a attack to solve various business problems. Our core operation is to facilitate car buying.
And selling and accurate car identification is very critical. A misidentification can cause friction with buyers and seller ad car sales. We have built various a attack to help us improve this process. For example, our award-winning Cyclops an AI assisted image, recognition tool eliminates approximately 55 hours of manual handling of around 20,000 daily vehicle images by caster. Stop Cyclops also improves the vehicle identification process for private and dealer customers, increasing accuracy and speed no manual lookup usage as photographed the back of
The car to identify it, we have also built a vin decoder called Navy which complement cyclops, who identified the fuel and transmission of a car Cyclops and movie, were both built using AWS, deep learning, ami running on GPU machines and AWS ECS cluster. This allow us to focus more on building the attack without worrying about infrastructures and GPU driver installation which save us a considerable amount of time. Car still received thousands of new colistin per day. Our customer service team manually checks every single ad to ensure our quality standards are maintained to help
Automate some of these manual checks we build AI, called Tessa since deployment of Tessa our operation efficiency has increased by approximately 50 %. An important role of Tessa is object. Detection, for example, radial number plates in uploaded photos. We use aw Sh maker object, detection algorithm with the help of AWS machine learning, solution lab team to build this technology, which only took us three weeks from proof of concept to production. This is possible because a WH maker
Eliminates the need for us to set up the AI framework building training code, set up, distributed, training cluster and build Insurance API, so letting us just to focus more on preparing a clean training set being able to deliver technology to squid is really critical to our team. So we can solve more business problems in order to protect our buyers. Identifying fraudulent inquiries is vital. Currently we use a number of different detection systems. One that is still in trial is a wh maker, placing text classification, another AWS
Offering we are considering is grant truth and reinforcement, learning which we believe will make our process even more accurate and efficient. Now that was a team at consoles, describing their amazing machine learning journey on AWS to continue learning why customers are choosing to do their machine learning on AWS. Please visit the demo arena at this conference to watch how machine learning is used in real-life applications. Now, if you have any questions and want to get them answered by AWS, experts head to the Asti experts area for any further information about ML on AWS, do visit the machine.
Learning on AWS page, as per the link shown on your screen now for those of you interested in gaining more confidence and hands-on experience with AWS, we encourage you to access the digital training built by AWS. Experts do attend our instructor-led classes by qualified AWS instructors and learn how to design, deploy and operate highly available, cost-effective and secure applications on AWS, validate your technical expertise with AWS and use practice exams to help you prepare for
Aws certification now wed like to thank you again for attending the conference, and we really appreciate your feedback as it allows us to better understand the topics and services you want to know more about. So please do take the time to fill out our survey and let us know what you think. Lastly, we request you to please watch on and more importantly, build on.
Learn more about AWS Innovate Online Conference at – https://amzn.to/2VihDMQ
AI and ML is the new normal. This session is intended as an introduction to how AWS customers are building intelligent business systems powered by AI and ML. Learn how these services, in conjunction with the large number of complementary AWS technologies, provide a great platform for our customer to build their own AI and ML powered solutions and drive business value. Towards the latter part of this session, hear directly from our customers about their AI and ML Journey on AWS.
Speaker: Barnam Bora, Head of AI & Machine Learning, AWS, APAC
Transcript
Hi, my name is Barnum Bora and I am the head of artificial intelligence and machine learning for the asia-pacific region. Here at Amazon Web Services, we are very thrilled to have. You join us today for the AWS innovate online conference, machine learning special edition 2019. Now this particular session building business outcomes with AI and machine learning is intended as an introduction to how our customers are building intelligent business systems powered by AI and ml, and how these
Services, in conjunction with the large number of complementary AWS technologies, provide a great platform for our customers to build their own AI and ML powered solutions to drive business value towards the latter part of this session. We will also hear directly from one of our valued customers about their machine learning journey on AWS in our industry. A very familiar and commonly used statement is that AI is the new normal and Amazon weve been making huge
Investments in machine learning since 1995, many of the capabilities, our customers, experience within amazon com and AWS are driven by AI. So for us, AI has always been normal, but this statement also holds quantitative, meaning on a much broader context. Lets dive deep to understand why this statement holds true and why organisations, both big and small public and private are rapidly accelerating their investments into AI in machine learning by multiple estimates, the gross domestic product of the world in 2017 is
Known to have been approximately 75 trillion dollars now this is a measure of the overall economic productivity of the entire human race. Many social and analytical organizations have made growth projections for the GDP to increase rapidly over the next decade and beyond. Technological innovation has been a key enabler in the ever accelerated increase of human productivity in recent history and technological innovation stemming from ail ml, is starting to take center stage across industries and across geographical boundaries. A PwC study recently estimated that global GDP will
Grow rapidly to approximately a hundred and twelve trillion dollars by 2030, this is clearly a great testament to the innovative efforts of our generation, but the same report estimates an even more exciting statistic that is of significant importance to our discussion in this session today. It is projected that about fifteen point, seven trillion dollars is going to be added to the global economy in 2030 by the productivity improvements and net new
Value generated by AI and machine learning, in other words, 14 % of total human productivity by 2030, is going to come from AI. Our customers realize the significant improvement and empowerment that AI can provide to them and their customers. Ai technologies, like computer vision, is helping. Doctors detect and prevent harmful diseases much earlier and more accurately than previously possible. Real-Time transcription and translation is helping bridge cultural and language barriers and make the world more accessible at AWS. We are striving to empower the builders who are embracing AI. Now. Our mission at AWS is
To put AI and machine learning in the hands of every developer, democratizing AI to all of our customers is a core mission for us, and we are seeing that just like how AI n ml has helped Amazon build a better business. Our customers are benefitting rapidly from adopting AI quickly. We are actively working to make sure our platform is accessible and helpful, not just to the seasoned ml practitioners and data scientists, but is also highly relevant.
And useful for emerging AI practitioners and developers now we have been using machine learning to power our business for over two decades. Many of the systems and evasions that our customers now rely on and experience everyday are powered by ml. The amazon com recommendations, engine which has been a core part of our retail business for many years, is driven by machine learning. Our Amazon, robotics division uses machine learning and autonomous robots to determine optimal picking routes within our fulfillment centers. In fact, our entire supply chain, forecasting and capacity planning are informed by machine learning. Algorithms
Now, beyond those use cases ml enriches our customer experience in new and exciting areas of innovation, such as our drone delivery initiative. We call prime air or our voice based products such as Amazon, echo, which leverage Amazon Alexa services and the computer vision and sensor technologies power. A revolutionary retail experience in amazon co stores, where customers can just walk out of a store without having to wait in a checkout line. Now all of these efforts would not have been
Possible without machine learning and deep learning, in fact, today we have thousands of Engineers at Amazon, committed to machine learning and deep learning and its always been a big part of our heritage, but driving business value with ml and the innovation with ml is an iterative process. Every day we apply new AI machine learning, based improvements to the Amazon business at a global scale through AWS. Now one such recent improved is the edition of Amazon scout Amazon Scout is a completely new. Take on the recommendations. Engine
Recommendations typically are based on data about past purchases, user behavior, similar transactions, correlations, etc, but some of our customers want a heavily visual buying experience where they arrive at their desired items through a visual pathway of preferences. Amazon Scout is an engine that uses computer vision, algorithms to recommend products based on visual similarity and preferences. We encourage you to go: try
It out the same computer vision, technologies and innovations that are used to build experiences like Amazon Scout and Amazon go check out. Less stores are available for our customers to utilize through the AWS platform, now working closely with our customers and learning iteratively from our own experience. Over the years, we found that there are some chronic challenges that have historically prevented organizations from adopting AI and machine learning quickly. We have consistently observed that ml expertise has been historically rare, building and scaling. These ml systems has been hard. Deploying and operating models has been
A time-consuming exercise and has typically been cost prohibitive. Hence, while building our services, our engineers actively try to address these roadblocks and mitigate them. The overall platform needs to be cost effective, easy to use and scalable in order to truly help our customers. Customers have also consistently shared with us that even for seasoned practitioners and data scientists, tasks like data acquisition, labeling framework selection, scale, our training model, tuning testing simulation and inferencing at scale has always been a
Challenging list of tasks, another consistent requirement both for our engineers and our customers, is that an ml solution would require training and inferencing, both in the cloud and on the edge and the experience needs to be seamless and manageable. This is one of the core capabilities we have engineered into the platform, especially with services like Amazon sage maker, taking
This feedback and learnings as design inputs and iterating consistently the AWS AI nml platform, is designed to be accessible to every developer and the services are designed to simplify and mitigate the key roadblocks and provide developers with the capabilities that helped them build. The intelligent systems that can empower their organizations, the AWS, AI and machine learning stack, is designed with three layers: AI services, machine learning, services and ml frameworks and
Infrastructure the layers correspond to the different levels of customization and configurability that different business solutions may need the top layer comprised of our AI services, which are plug-and-play API services that offer instant ml based outputs in the areas of vision, speech, language chat, BOTS forecasting and recommendations Developers can call these apis directly and integrate the AI functionality they provide directly into their solutions without the need to explicitly learn and become data scientists or AI
Practitioners, these AI services, are used by tens of thousands of our customers daily for a broad range of a I backed applications, especially the ones that require human interaction or repeatable. Similar acts like recommendations now the second layer contains the plethora of ml services and functionality. That is provided to ml practitioners through our Amazon sage maker service for situations
Where our customers want to build fully customized ml models and deploy them at scale, Amazon sage maker is the fully managed ml platform that provides the complete end-to-end ml functionality from building to training, to testing and then deployment and management. The sage maker platform greatly simplifies and accelerates the process of building ml powered systems from scratch by providing a large amount of inbuilt functionality and manageability across all major stages of machine learning. Development now for solutions that require full configurability of the
Infrastructure where customers need to utilize many other open source technologies. The third layer provides the most comprehensive list of compute targets and infrastructure options, both in the cloud and the edge to truly build end-to-end ml based solutions. Our customers have the option of utilizing CPUs, GPUs, FPGAs containers, serverless, architectures, elastic inferencing or even custom design, inference hardware like AWS influenza and even edge devices through AWS, IOT and Greengrass. Now it is important to remember, though, that
Machine learning is only a component of the overall business solution and the depth and breadth of the broader AWS services provide the true end-to-end capabilities that an organization needs to be successful in their ml journeys. Aws provides a vast and comprehensive platform that our customers know and love, and the entirety of the platform greatly extends the capabilities of the ML solutions. Customers want to build now a common
Pattern we observe across our customers and our internal business teams in Amazon is that driving business value with a line. Machine learning is a cyclic process that involves getting relevant data and applying domain knowledge and expertise to that data to build intelligent business systems and experiments that provide business value. This business value typically manifests itself in five broad categories of outcomes, typically systems and processes that involve a lot of repeated decisions of similar type like fraud, detection,
Systems etc are being automated by our customers using machine learning. Machine learning based signals like in the case of weather forecasting systems are used by professionals to make critical decisions like planning airline routes. Now forecasting is a perfect example of a type of business task where predictive ml is helping in looking at future behavior. Our customers are prescribing major process changes as well from ml based intelligence and simulation. That is helping them greatly, improve their productivity.
Now, finally, we have also observed that a lot of customers identify net new capabilities to further their businesses. For example, our customers are utilizing technologies like computer vision in self-driving cars to disrupt their industries. Our customers are actively building innovative ml based intelligence systems to empower their businesses at a global scale through AWS. Now let us look at a few of them to see what they have been able to achieve with the platform into it. The builders of the
Popular accounting software QuickBooks have been able to accelerate their ml development. Utilizing Amazons sage maker, resulting in a 90 % reduction in deployment time time to innovation and value, is a game-changing capability for most organizations, and customers, like Intuit, have taken full advantage of the acceleration that the platform provides. Our customer to simple are a self-driving technology. Startup company based out of China, who have utilized the AWS ML capabilities to build a revolutionary self-driving platform that can plan the driving route, a thousand meters ahead, while on the road
Now, using Amazon sage maker Formula Ones, data scientists are training deep learning models with more than 65 years of historical race data to drive insights that they never had access to. Before now, bgl corporate the builders of the popular fund management software, simple fund 360, have built a fully AI based fund management assistant that utilizes deep learning our customer Sunday insurance, who are an inshore tech startup in Thailand, have built a fully a I based insurance underwriting Engine
And are disrupting the insurance pace across the region, car sales, calm, Australias number one car marketplace: utilize: the AWS AI nml platform to do automated image, classification and multiple other ML back systems to build a better experience for their customers. Now tens of thousands of companies are well and truly on their journeys. Utilizing the AWS, AI and evelle platform, the completeness of the platform and the flexibility and reliability of AWS are
Fundamental to why our customers are choosing to build their ml solutions on AWS now lets hear from one of our valued customers car sales about their ML journey with AWS. I am a Christmas nellwyn at the head of AI at Carsons. Our teams role is to research and build a attack to solve various business problems. Our core
operation is to facilitate car buying and selling and accurate car identification is very critical a misidentification can cause friction with buyers and seller ad car sales we have built various a attack to help us improve this process for example our award-winning Cyclops an AI assisted image recognition tool eliminates approximately 55 hours of manual handling of around 20,000 daily vehicle images by caster stop Cyclops also improves the vehicle identification process for private and dealer customers increasing accuracy and speed no manual.
Lookup usage as photographed the back, of the car to identify it we have also built a vin decoder called Navy, which complement cyclops who identified the fuel and transmission of a car Cyclops. And movie were both built using AWS, deep learning, ami running on GPU machines and AWS ECS cluster. This allow us to focus more on building the attack without
Worrying about infrastructures and GPU driver installation which save us a considerable amount of time car still received thousands of new colistin per day. Our customer service team manually checks every single ad to ensure our quality standards are maintained to help automate some of these manual checks. We build AI called Tessa since deployment of Tessa. Our operation efficiency has increased by approximately 50 %, an important role of
Tessa is object detection for example radial number plates in uploaded photos we use aw Sh maker object detection algorithm with the help of AWS machine learning solution lab team to build this technology which only took us three weeks from proof of concept to production this is possible because a WH maker eliminates the need for us to set up the AI framework building training code set up distributed training cluster and build Insurance API so letting us just to focus more on preparing a clean training set being able to deliver technology to squid is really critical.
to our team so we can solve more business problems in order to protect our buyers identifying fraudulent inquiries is vital currently we use a number of different detection systems one that is still in trial is a wh maker placing text classification another AWS offering we are considering is grant truth and reinforcement learning which we believe will make our process even more accurate and efficient now that was a team at consoles.
describing their amazing machine learning journey on AWS to continue learning why customers are choosing to do their machine learning on AWS please visit the demo arena at this conference to watch how machine learning is used in real-life applications now if you have any questions and want to get them answered by AWS experts head to the Asti experts area for any further information about ML on AWS do visit the machine learning on AWS page as per the link shown on your screen now for those of you interested in gaining more confidence and hands-on experience with.
AWS we encourage you to access the digital training built by AWS experts do attend our instructor-led classes by qualified AWS instructors and learn how to design deploy and operate highly available cost-effective and secure applications on AWS validate your technical expertise with AWS and use practice exams to help you prepare for AWS certification now wed like to thank you again for attending the conference and we really appreciate your feedback as it allows us to better understand the topics and services you want to know more about so please do take the time to fill out our survey and let us know what.
you think lastly we request you to please watch on and more importantly build on