Machine learning has graduated from the realm of science fiction to become a core, transformative technology for organizations across industries and categories. The unique potential and power of machine learning is sparking genuine innovation, powering the ideas that are improving lives and protecting our planet right now. With machine learning, organizations are making inroads toward ending the pandemic, protecting and supporting our veterans, finding homes for the homeless, understanding climate change, and more. But this is just the beginning.

“The technology is ripe, and it now has the ability to provide new and significant solutions for some of the world’s biggest issues,” says Michelle Lee, vice president of the Amazon Machine Learning Solutions Lab.

Tens of thousands of companies and organizations worldwide have turned to Amazon Web Services (AWS) for machine learning – from BlueDot, who is tracking disease outbreaks worldwide, to the Fred Hutchinson Cancer Research Center, finding new ways to treat cancer, and Mantle Labs, a startup using machine learning to offer farmers cutting-edge crop monitoring, and more.

However, access to machine learning, a new technology to so many of these organizations, can often come with a skills and technology deficit. That’s where AWS steps in, partnering with innovators to bridge the gap and bring pioneering solutions that tackle our most urgent and important challenges.

Helping companies develop groundbreaking machine learning solutions

The Amazon ML Solutions Lab pairs an organization with machine learning experts to help identify and build machine learning solutions that tackle the issues at the heart of the customer’s mission and purpose. Whether you are a large global enterprise, a startup, or a non-profit, the ML Solutions Lab brings over 20 years of Amazon ML innovations to help organizations get started with machine learning.

“A lot of these customers are addressing new opportunities where they’re looking at more exciting and more efficient ways of doing their business or going after research problems that were previously untenable,” Sri Elaprolu, senior manager in the ML Solutions Lab explains.

In designing a machine learning solution, Elaprolu says it’s essential to consider the capabilities and skill set that a particular organization has in-house to operate the solution long-term – and not hand off something that is beyond the abilities of the organization to maintain.

“As part of the Lab we have a global mission,” Elaprolu continues. “What we’re doing is impactful in the real world for everyday people, and the results are extremely compelling. Applying technology to solve real-world problems that have a meaningful impact on humans and life is absolutely thrilling.”

Let’s take an in-depth look at some of the most important work being done today.

Since 2012, RallyPoint, a social media platform designed for the broader U.S. military community, has provided an online user experience focused on military service members, veterans, families, caregivers, and survivors to help them lead more successful and fulfilling lives. Among the millions of public discussions on the platform, a small percentage come from members who share thoughts and behaviors about self-harm. The Department of Veterans Affairs estimates that approximately 17 military veterans die by suicide each day – and RallyPoint has made it a priority to offer critical mental health resources and support to these men and women when they need it.

Developing a way to quickly and accurately sift through these high-risk public posts created by a small minority of RallyPoint users is a challenge. In order to speed discovery of these at-risk public posts, RallyPoint turned to the ML Solutions Lab. The Lab worked closely with RallyPoint to develop a machine learning model that can quickly analyze public posts on the RallyPoint platform and help determine whether there is an indication of self-harm. With the help of this machine learning model, RallyPoint has been able to successfully flag concerning posts quickly and accurately while reducing the amount of manual review needed to enable a potentially life-saving intervention.

Behind the solution

The team at the ML Solutions Lab also worked closely with RallyPoint and mental health experts at Harvard University’s Nock Lab to tackle this challenge. First, the Amazon ML Solutions Lab collaborated with RallyPoint to build a machine learning model using Amazon SageMaker and anonymized public posts provided by RallyPoint. Then, mental health experts at Harvard helped train the model by annotating additional posts using Amazon SageMaker Ground Truth in order to continuously improve the accuracy of the predication made by the model. Ongoing, RallyPoint and Harvard will continue to further refine the model while evaluating the best content (e.g., mental health programs, hotlines, support groups) and preferred method to surface information to users. In the long term, the goal of the solution will be to augment the community engagement by RallyPoint member administrators that takes place on the platform today when there is self-injurious content.

“We are encouraged by the early results – and how the technology is contributing to tackle this challenge,” Lee says. “It is our privilege to support the military community in this work.”

As of July 2020, COVID-19 has infected more than 17 million people worldwide, and more than 674,000 have died.

Since the virus was first identified in late 2019, a huge amount of cutting-edge research on ways to fight COVID-19 has been published, with more appearing every day – coming so fast that researchers can’t keep up with it. Faced with an exponentially increasing volume of information, world researchers are finding it difficult to derive insights that can inform treatment and prevention.

To help combat the problem, the Amazon ML Solutions Lab worked with teams across AWS, to build and launch CORD-19 Search, a new search website powered by machine learning that can help researchers quickly and easily search for research papers and documents.

How CORD-19 Search works

CORD-19 Search was built on the Allen Institute for AI’s COVID-19 open research data set of more than 130,000 research papers and other materials. This machine learning solution uses Amazon Comprehend Medical to extract relevant medical information from unstructured text, including disease, treatment, and timeline. The information is then indexed in Amazon Kendra, an enterprise search service with natural-language query capabilities that make it easier to find and rank related articles.

The platform returns the most relevant articles corresponding to a researcher’s question, along with other related materials that may be of interest. The research can now be focused in a much more narrow, relevant space, versus having to sift through thousands of results.

To help researchers find and visualize insightful relationships between scientific articles, the team introduced a knowledge graph. The COVID-19 knowledge graph incorporates articles, authors, institution affiliations, citations, and biomedical entity relationships; the resulting graph contains over 336K entities and 3.3M relationships. Perhaps most helpful to researchers, the knowledge graph powers a recommendation engine, surfacing the most highly relevant articles based on a user’s search query and even browsing history. And the full knowledge graph has been made publicly available to researchers through the AWS COVID-19 Data Lake to enable future insights and discovery.

“AWS’s long-term vision is to expand the CORD-19 Search architecture to incorporate even more data resources than what we’ve already incorporated,” Lee says. “This will allow researchers to uncover patterns of disease progression, make data-driven decisions, and help improve patient outcomes in the effort to unlock data related to COVID-19.”

AWS knows that the next brilliant innovations may just now be percolating in the minds of brilliant, civic-minded entrepreneurs and engineers who need assistance to bring their ideas to life. They’ve invested in developing resources to support those looking to use machine learning in this way, beyond the Amazon ML Solutions Lab. These include programs like the AWS Imagine Grant Program and the Amazon Research Awards.

AWS Imagine Grant Program: Helping PATH change lives in L.A.

The AWS Imagine Grant is awarded to non-profits and non-governmental organizations that are using powerful technology to solve some of the world’s toughest challenges. It provides grant winners financial and operational support including AWS Promotional Credits, training services, marketing support, and more.

Recipients are finding cures for childhood cancer. Stopping illegal fuel dumping in our oceans. Sharing knowledge and culture. Giving unbanked populations a financial voice. Guiding women facing life-threatening breast cancer diagnoses. Helping veterans access the support they’ve earned.

Or, like PATH – another recipient of the AWS Imagine Grant Program – using machine learning to address homelessness.

For someone living on the street or in a homeless shelter in Los Angeles, the wait to get housing through the county’s Coordinated Entry program usually takes months.

PATH, an L.A.-based organization founded to address the ever-increasing issue of homelessness, applied for an AWS Imagine Grant to develop a way to shorten that wait dramatically. With the grant and support from the AWS team, the organization developed LeaseUp to connect clients in real time with the best possible housing for their needs.

Amazon Personalize captures relevant information about available units of housing so case managers can recommend the best housing option to their clients in real time. By integrating this technology, the organization has been able to match over 600 individuals experiencing homelessness with housing – and reduce the time it takes to do so. Timing in these situations is often critical; a person who is ready to come in and get help one day may not return the next.

LeaseUp aims to add 2,000 new units to its database over the next year to help even more people make it home. Bringing more existing apartments onto the platform, as well as working more seamlessly with the landlords to list rental units, are important steps in not just addressing homelessness but in ending it.

The Amazon Research Awards are dedicated to creating the future with scientists around the globe – and a key component of that mission is helping academics advance the frontiers of machine learning.

Providing access to the latest compute, storage, and networking is key to lay the groundwork for PhD candidates and graduate students to further their research. Now, recipients of the award are using machine learning and its applications across a wide range of problems, from finding new therapies for cancer to solving climate change and exploring outer space.

The awards provide eligible researchers and university programs with cash awards and AWS Promotional Credits so that they can do more, more quickly, using the most advanced compute, analytics, and machine learning tools available in the cloud.

Climate scientists at Oxford are working to unearth new ways to combat climate change, as a recipient of the Amazon Research Awards. Machine learning is an essential tool for climate change research since climate science is such a data-intensive field. Climate models are enormous, requiring supercomputers to run them, and analysis requires a huge amount of earth observation data. As data continues to grow, along with complexity, it becomes impossible to explore all avenues of research manually. This is one of many ways AWS and Oxford continue to work together, including a recently announced collaboration to fund a testbed of new research in AI and data science across the university.

In the Climate Processes group in the Department of Physics at Oxford, the hope is that by studying the effects of aerosol pollution on clouds they’ll be able to break new ground in global warming research, leveraging tools like Amazon Deep Learning AMIs running on EC2. Clouds reflect sunlight back to space, acting like an umbrella that cools the earth. Hence, even small changes in clouds in response to global warming or air pollution could have a big impact on environmental health and serve to accelerate or dampen the greenhouse gas effect. Machine learning models can track these changes to understand why clouds change, which could be the key to addressing global warming.

Now Oxford scientists will be able to analyze satellite data covering the entire earth multiple times a day, providing countless images of aerosol-impacted clouds, which they’re able to process in the AWS cloud, thanks to the grant from AWS. In September, 15 PhD students across Europe will start working with teams at Amazon to train on the machine learning tools that will help quantify these effects, and understand their dependence on cloud type, which regions they form in, where they are, and how prevalent they are. “Such scalable machine learning techniques allow us to make rapid progress in an area where researchers previously spend months of their time on identifying features in fairly limited datasets manually,” says Philip Stier, a professor of atmospheric physics.

“We’re working in a field that is fast-emerging, alongside a team of highly accomplished and experienced scientists and pushing the boundary on a daily basis,” says Elaprolu. “A lot of the problems that our team tackles have not been dealt with previously. We’ve seen the power of machine learning when applied, and how transformative it can be.”

Organizations are constantly working on innovative techniques to solve the most important issues the world is facing today, making profound and significant impact across the world. And the Machine Learning Solutions Lab is there, bringing its technical skill and expertise to its customers, helping them pursue their world-changing goals.

“The greatest transformational impact comes when we bring together technology experts, such as those we have at AWS, with our customers’ subject matter expertise,” Lee says. “When you combine those two, we have the potential to create powerful change to build for a better today.”