Building neuroinclusive cloud experiences

Hello, everyone and welcome to this talk on building neuro inclusive experiences.

Thank you so much for coming in today. Before I start off, I love knowing the audience a little bit more. So could we have a quick show of hands on how many folks in the room have heard the term neurodiversity before this? Oh, that's wonderful. So then probably my second question becomes redundant, which was going to be how many have never heard the term neurodiversity or neuro divergence? Good.

So it might literally come across as me preaching to the choir, but I'm going to do the best that I can. My name is Manisha Mehta, my pronouns are she/her and I am the Global Disability Inclusion Leader for AWS and I'll be taking you through today on what is neurodiversity? That's what we're going to be covering over the next 15 minutes.

I'll also share some insights and statistics around neurodiversity - how is it, how does it land in the workplace? What are the challenges that exist around this topic and how can we overcome them? I'll also share a quick sneak peek of how we in AWS are landing neuro inclusion. So what are those activations or steps or initiatives that we are taking to make AWS more neuro inclusive.

And then I will hand over to Tash who is going to be presenting with me today, Natasha McCann, on how to build neuro inclusive services and solutions for our customers, for you.

So let's jump right in. A little bit about me to start. So about seven years ago, I started deep diving into the subject of disability and neurodiversity inclusion. I interacted with a lot of members from the community. I read a lot of articles and I learned about the misconceptions, the challenges and the barriers that the community faces around the world and in the workplaces.

About a year ago, my daughter, who was then five years old, was diagnosed as being on the autism spectrum. And I think today the work that I do helps me enable her a lot better in creating a more neuro inclusive workplace and a neuro inclusive society. And her diagnosis has actually helped me understand my own traits a little better as well.

I grew up in a country and at a time where neuro divergence was not understood and it was not formally diagnosed as a condition. But now thanks to her, if I look at myself and the fact that at any given point in time, I have 20 tabs open, I'm reading three books and working on at least three different projects or documents or events at the same time, classic ADHD traits. So it's probably thanks to her that I'm getting them, at least understanding them and navigating them a lot better.

So with this context, let's understand what neurodiversity actually is. Neurodiversity as a term was coined by an Australian sociologist, Judy Singer in 1998. And she came across this term and she coined this term when she started noticing similarities between the social traits of her mother and her daughter, both of which did have challenges around non-verbal communication, around their social skills.

And later Singer's daughter was, was diagnosed as being on the Asperger's syndrome, which is now merged into the autism spectrum disorder. So we now do not use the term Asperger's anymore, but she's the person who actually first coined the term neurodiversity to land the fact that neurological differences are a part of the diversity within this space and they are natural variations in how our minds think, act and how we communicate.

So why are we calling this a spectrum? We call it a spectrum because it considers both neurodivergent and neurotypical individuals. When we say neurodivergent, we are actually talking about 20% of the world, 20% of the global population identifies with neurodivergence, which means that their mind, their executive functioning, their cognitive skills is different from what a neurotypical person thinks, acts and communicates, which as you can extrapolate is about 80% of the population.

Neurodivergence, as I've mentioned before, is primarily genetic. It can also be acquired, but that's in very, very rare cases. And I can't resist the pull of sharing some famous names with you, which you may have heard before - Albert Einstein, Tom Cruise, Emma Watson, Justin Timberlake, Steven Spielberg, Bill Gates and Richard Branson.

Richard Branson actually calls out dyslexic thinking as a superpower, as a skill on his LinkedIn profile. So there are probably more neurodivergent thinkers that you work with and that you are probably interacting with than you are aware of.

When we look at neurodivergence or the spectrum conditions, it actually is formed of a set of a number of conditions under that. And you may have heard across some of or have come across some of these terms:

• Dyslexia - when a person has trouble reading, understanding spellings. So basically written or verbal communication.

• Dyspraxia - when somebody may have a condition where their hand-eye coordination is noticeably lesser than what an average hand-eye coordination looks like. So we do tend to label some people as clumsy, it could be because they are dyslexic.

• Dyscalculia - again, something with someone like me has trouble with math, with formulas, with long calculations that again comes under neurodivergence.

And then the more often heard names of autism spectrum disorder and attention deficit and hyperactivity disorder.

I'm going to go one level deeper into what an average neurotypical cognitive profile looks like, just for us to understand how neurodivergence shows up and what does executive functioning mean.

So as you can see, the gray band on this slide is the average range. So for most folks who are neurotypical, 80% of the population, our skills are going to be lying within this gray bar that you see there. And when I say skills again, I don't know if you can read on the screen - verbal analogy, vocabulary, skills, reasoning, spatial awareness, just visual memory - all of these skills come under the cognitive skills spectrum. And this is what a neurodivergent neurotypical profile looks like.

I'm now gonna share a dyslexic cognitive profile. But even before I share or talk to this profile, I want to start with a quote by Dr. Steven Shore who says "When you meet one person with autism, you've met one person with autism," which means that the profile I'm showing on the screen right now is a single person's profile. It doesn't hold true for every single person who is diagnosed as being dyslexic.

So what does this profile tell us? This profile actually sometimes is also called a spiky profile. And I think it's very obvious why - this particular person is 96 percentile on their spatial awareness. They are going way above the average range, which means that they're going to be very, very good at collating and working with abstract data and presenting that in a visual model or a design. That's going to be their superpower, right?

And then you can see that there is verbal memory on which they are at 12 percentile, which means when you're communicating with them, you will see certain challenges that they face. And this is how neurodivergence shows up. And this is why it's called a spiky profile - there are some skills where I'm going to be really, really high, which is going to be my superpower and some where I may need some adjustment or accommodation.

The other thing I want to highlight today is the fact that 80% of disabilities - mental, physical, neurological - are not actually visible. So unless I share with you that I could potentially be ADHD, when you look at me, you're not going to be able to make this out. And that's very important when we work in this space or we start discussing neurodiversity or disability inclusion, because when I am not able to see a particular condition which could be limiting for an individual in terms of presenting a challenge, then I kind of go down the road of misperceptions, biases and making assumptions and labeling somebody, right?

So it's really, really important to keep in mind that we need to create a better understanding amongst ourselves to become more inclusive and not label persons without knowing what kind of lies behind their behavior or their condition, right?

So I did mention that I'm going to share some insights, statistics on why neurodiversity makes business sense. And then of course, a sneak peek into how we in AWS are addressing this.

I have selected these statistics very intentionally because I did want to bring out the biases that neurodivergent individuals do face today. As you can see in the first bullet itself, only 16% of autistic individuals actually are in full time employment. And data tells us that this is simply because of the lack of understanding and awareness that they're getting in the workplace, which is actually reflected in the second bullet - 43% of folks, so imagine only 16% get jobs in the first place, but 43% of these 16% are actually losing their jobs because they are not getting the accommodation they need, they are coming across as too direct, as too literal, their social interactions may not be as we expect it to be. And hence they are losing their jobs sooner rather than later in terms of their tenure.

And which I think also brings across the fact that of that 12%, the fact that only 12% employers are actually conducting trainings or awareness sessions around neurodiversity inclusion - 88% of us need to do a lot better than what we are doing.

And very quickly to land some statistics on why neurodiversity makes business sense - with increased automation, with increased digitization, the need for traditional skills around reading skills, memory abilities, coordination, time management is decreasing, which are typically associated with neurodivergent challenges. And what is trending on the other hand is what is a neurodivergent strength - creative thinking, analytical thinking, lateral thinking, thinking outside the box, eye for detail.

So all of these trending skills are neurotypical strengths and hence they absolutely make business sense for us to create teams that are neurodivergent and neuroinclusive. And this research, as you can see, this quote is from Harvard Business Review, but you can come across research from EY, from Accenture who have a lot of data around the business benefits that neurodivergence presents.

The research that was conducted showed us that 80% of individuals who are neurodivergent are more hyperfocused, 78% are more creative, 75% showed innovative thinking, 71% have an eye for detail - they are able to work with large data patterns, understand patterns inside large data streams. And Natasha is going to talk more about that from her own personal experience later in the presentation. But there is an absolute business case around why neurodivergence makes sense.

And lastly, what are we doing and how can we create real neuroinclusion in the workplace? The answer to this, the key to this is surprisingly simple. The one thing that you need to keep in mind is the fact that we need to limit ambiguity. We need to give clear instructions, set very, very clear goals and communicate clearly around what are the behavioral and performance expectations we have from somebody in the role that they are doing.

We need to plan ahead to avoid surprises. I mean some of us love surprises, some of us don't. But again, I would love to share a personal anecdote here with my daughter - the plan ahead bit really works when we communicate with her any upcoming change that's likely to happen, whether it's in her classroom, her teachers, a family trip, my work travel, the sooner I communicate to her, the more time she has to adjust around this and work it around in her brain and life is fine. But if there is something that comes across as a complete surprise for her, you can imagine the tantrums that happen and it does land for us as adults as well. Anyone who's neurodivergent, try to avoid surprises as far as you can in the workplace.

Communication is the other thing - be clear, avoid jargon, don't expect the person to read between the lines, just spell it out and say what exactly you're trying to say in terms of the goals or in terms of expectation.

And lastly, keep in mind the fact that individual needs, unique needs will always vary. There is no one size fits all.

It's always good to normalize empathy and ask that one question that says, "How can I support you in delivering your best work?" If there is one key takeaway that I would love for you to take, it's this one line: "How can I support you in delivering your best work?" That's all you need to do.

Now on the last slide on how we are landing neuroinclusion in the workplace. So earlier this year, we launched the inaugural disability and neurodiversity strategy for AWS. And we have focused our work around these four pillars that you see on the slide:

1. The first focus has been assessing our talent management processes, the benefits that we offer our employees, what is the talent review process that we have, how neuroinclusive it is, and then given the feedback that we get through our assessment, we are going to start working on our 2024 strategy plans.

2. The second bit of work has been around reviewing the way we recruit individuals. So are we communicating very clearly what the job description of the role is like? Is the goal spelled out? Are we keeping the sound sensitivity and the harsh lights in mind when we are conducting the interview process itself? Because that's again, you have to make the candidate present their best self and you have to facilitate that for somebody who is neurodivergent.

3. Our third area of focus has been around shifting the narrative on this topic. So conversations like this one around what is neurodiversity training? All of our managers, our HR teams and all of our employees, our builders are trained on what is neurodiversity, how do you work around it, how do you create the adjustments and accommodations that are needed - not just to make AWS more inclusive, but I think also to create solutions for our customers that are neuroinclusive, that are built into the designs and the coding and the solutions that we are doing for all of you.

4. And last, but not the least, we rely on our remarkable community of 18,000 inclusion ambassadors. And if you want to learn more about it, I'm going to be at the All Builders Lounge over the next couple of days. I would love to share more on this, but we do have 18,000 volunteers within AWS, our employees who are passionate about inclusion and they help us amplify the work that we do and spread the word around all the inclusion, diversity and equity topics and initiatives that we're running in AWS. So we do rely on this community again to spread the word on neuroinclusion and that's built into our strategy.

And with that, I'm going to hand over to Tash to take you through exactly how we're building neuroinclusively, and hopefully practically show you what neuroinclusion looks like.

Tash:

Hi everyone. My name's Tash. I'm a solutions architect at AWS. I help customers build on our cloud platform. Neuroinclusion and neurodiversity is an important subject area for me because at age 13, I was diagnosed as neurodivergent. For a very long time, I kept my condition a very closely guarded secret. But after someone very close to me getting diagnosed as well, I recognized that I needed to be more open about it myself and share my learnings with the public like I'm doing right now. So hopefully I can share some insights with you after nearly a decade in the tech industry. And hopefully you can take some stuff away with you back home.

I'm gonna be focusing on three key themes today on building neuroinclusively that I think are really pertinent in today's world and what's going on in the world at the moment.

1. The first one being responsible AI, how you can build AI systems, neuroinclusively.

2. Improving accessibility using machine learning. So being able to use machine learning in order to help solve different tasks.

3. We'd be finishing off with building diverse teams for stronger cyber defense.

So let's start off with responsible AI. First of all, who is using AI right now, who has used artificial intelligence when they've been developing? Fantastic.

So before we go into how you can build AI systems, neuroinclusively, I just want to touch on what the different elements are kind of within the AI landscape.

So what is AI? AI is a way to describe any system that can replicate tasks that previously required human intelligence. Almost all of these AI systems that we see today are created using machine learning. And machine learning uses large amounts of data to create and validate decision logic, which is known as a model.

You then have deep learning, which again is a type of machine learning. But you might see on the slides that that picture up there kind of represents a human brain. So deep learning really tries to replicate how the human brain functions which allows those systems to then address more complex use cases than was previously possible.

You've then got reinforcement learning. So if you've ever come across an automated vehicle, this uses a form of reinforcement learning. And this will essentially train models by using some kind of reward function created using code. If you want to see what this looks like for real, there's actually a DeepRacer stand in the Venetian Expo hall which kind of shows how reinforcement learning works in AWS. So go have a look at that if you're interested.

Who's heard of generative AI raise your hands? So this has really taken the world by storm in 2023. As I'm sure we've all guessed. And you may be wondering what is the difference between this and the AI we used to name. So generative AI focuses on creating new content while traditional AI solves specific tasks. And it's really taken the world by storm because we've been able to through consumer facing applications, applications that someone non technical could use has allowed them to create new content and ideas for themselves.

I just wanna touch on also some of the dimensions of responsible AI that we build on at AWS. So what really constitutes responsible AI is still being debated. AI is a new thing. It's gonna change over time. And at the moment, we kind of represent responsible AI as being made up of these six key dimensions:

1. Fairness - how we consider how a system affects different subpopulations of users, for example, by their gender, disability status, ethnicity or language. We want to make sure that there's not harmful disparities in how that system performs across those subpopulations.

2. Explainability - whether that system offers a clear rationale for its decisions. So not just the how and the what but also the why. We want justification for the inference and the output produced by that system, which is particularly important for those with compliance requirements. I mainly work with public sector organizations and that's very, very important to them.

3. Robustness - We'd also like confidence that the AI system cannot be easily fooled or confused and success here means at the least that we have very understandable guidelines to both stakeholders and to ourselves for when that system is expected to work.

4. Privacy and security - If you've been to an AWS presentation before, you may have heard the word security is the highest priority at AWS and that doesn't change whether it's AI or not. We really set guidelines for you to follow under what we call the shared responsibility model that help ensure and help you to ensure that your data is used in accordance with privacy considerations and also your compliance requirements.

5. Governance - This kind of slightly builds into the security. These are essentially processes and policies you set to enforce and ensure responsible AI practices are being used across your organization.

6. And finally, transparency - the extent to which an organization or that person building the AI system is transparent with both their stakeholders and sometimes the public, because AI systems can often be open to the public to use as well, to how that AI system is used - what data you should put into it and how you should interact with that system from a security perspective for example.

So now I've spoken about what AI/ML is, kind of what the dimensions of responsible AI are, now let's look at how we are building our services responsibly in terms of AI. And I'm gonna relate this a little bit back to the neuroinclusion part.

1. The first one being CodeWhisperer - you want a developer in the room? Fantastic. Raise your hands. Yeah. So this is an AI coding companion. CodeWhisperer essentially helps you write code using natural language processing. So I'd essentially write a comment in my interactive developer environment and say "Build me a DynamoDB database with a Lambda connected into it that can write into that database and that Lambda is connected into an API that feeds that code with data." And then CodeWhisperer will suggest code for me based off of that. So kind of relating this back to my experience as a neurodivergent person, I can often really struggle with processing information quickly. I actually find coding quite difficult because when I try to code with the syntax, those words and that syntax can often jumble around a little bit in my head, which is why I find it difficult. And I use now CodeWhisperer most of the time when I'm developing code or building proof of concepts for my customers.

2. The second one I'm gonna speak about is Amazon TITAN foundational models. So these are foundational models that we've created here at Amazon to help with different tasks like summarizing text, extracting text of a web page plus many more. So if you think of screen reader technologies here, I used to use screen readers in order to help me read web pages and also to digest the content on web pages, just to summarize text for example. And these foundational models are also specifically built to filter out any harmful content that may come back to me as well, kind of relating back to those responsible AI practices from earlier.

3. Another example of a service built with responsible AI in mind is AWS Healthscribe. If you've ever been to see a doctor, which I'm sure we all have at some point in our lives, and they've had one of those little recording devices - this is essentially what Healthscribe will operate on. That recording device will often output an MP3 or MP4 file for example, and you can essentially feed that into Healthscribe and get it to transcribe that file for you. So in healthcare, it's really important to have accurate medical notes. One in seven people in the world are neurodiverse - remember that doctors, clinicians, patients, they're probably at least one of them neurodivergent on each side of that conversation. They want to process that information in an accurate way, they don't want to be misquoted. So this really helps break down communication barriers and bias filters in healthcare. If someone is telling me something that might be slightly distressing to hear, I might not process that information properly. I might want to go back over those notes later on. So that's really helpful for me as a neurodivergent person. And I'm sure other people that are not neurodivergent too.

So now we've spoken about those services, I'm now gonna leave you some best practices that you can take away to build generative AI and AI systems responsibly:

1. Firstly, before you even think about using artificial intelligence, find your use case - figure out the why do you want to do this - and then focus on the how and the what. What do I want to achieve and make that as specific and as narrow as possible. The more specific it is the better the AI system will perform.

2. Prioritize education and diversity in your workforce. I've spoken on quite a few panels about neurodiversity and a lot of people through no fault of their own do not know what neurodiversity is. If you have a team of 20 people, at least two of those people on that team are neurodivergent. It's both a social and commercial imperative to make sure that those people and all of your team are supported.

It's so important to educate people on what neurodiversity is and how maybe different technology systems could help also assess risk with a performance evaluation.

It's great to assess how well your AI model or your code is performing. But what about the risk? It's just as important to evaluate those and do a risk assessment.

If you're a software developer, you might be slightly familiar with these words - test, test and test. Test how your AI model responds to different inputs and stimuli continually iterate. Don't just train your model once - iterate the model with new data and new insights over time.

Technology and data changes all of the time. We know how fast paced this world is. The AI model will probably need to change too.

Also introduce policies for governance. For example, have you got neurodivergent people on your testing teams in order to evaluate how your code is performing or to assess risks? Train your teams on those policies and measure those policies over time to see how those policies are performing or if the risk has changed.

So that's responsible AI. Let's now move on to how we can improve accessibility using machine learning.

So let's go into the definition of accessibility. First of all, accessibility is the concept of whether a product or a service can be used by everyone, no matter how they encounter that. It's not something that is either true or false. Accessibility is only measured in relation to a specific ability or a scenario.

For example, if you think of screen reader technologies that I was speaking about earlier, someone using that piece of technology in order to help them read a web page, if there isn't any issue with that technology, then that web page will then become inaccessible to that person using it. Which is why it's so so important that we build technology solutions with accessibility in mind.

If that web page is inaccessible to one person, it's inaccessible to all. So let's look at a few other examples:

• In direct relation to accessibility, you can use machine learning to extract and analyze data from documents, helping those who may have learning difficulties such as dyslexia, help interpret large bodies of text.

• Computer vision can help those who may be visually impaired to help navigate complex environments safely, for example, informing them of risks or obstacles and even help again as screen readers.

• Automated driving - while this is still a very early concept and not quite accessible to everyone yet - if we think of the future and where this could go, driving can be a complex task. I often get a little bit stressed out driving on a motorway because it's really loud, there's lots of things going on around me, and a lot of people with disabilities are unable to drive because of this. So if a car was built where you could simply press a button and go, that could really enable someone with a disability to have access to that vehicle, essentially having like a personal taxi on demand.

And I'm now going to go into a demo showing you a few examples of how AWS services can be built to improve accessibility.

So a great colleague of mine put together a blog which you can go and read. I've got a few QR codes that you can scan that showed how to create an application called AWS AugmentAbility. So this uses a combination of different machine learning services to achieve different outcomes. For example, transcription, translation, also identifying objects.

So let's go into how this was built in the architecture:

• First of all, you access the application for a mobile client, that app is coded in the SDK for JavaScript.

• You authenticate to the application using Cognito and that can be configured either publicly or privately depending on how you want to use it.

• That website is hosted on AWS Amplify and Amplify gives developers a prescriptive framework to build and host a web page.

Once authenticated, you can then use a combination of one of these machine learning services on the right to achieve your tasks.

So let's see how this looks for real with a few examples.

The first one I'm going to show you is quite similar to that HealthScribe example I was talking about earlier. This app will essentially run in the background on your mobile device and listen to what you're saying and then transcribe that for you. And this is done using a completely serverless application called Amazon Transcribe.

If you wanted to switch it around the other way and type something in and get it to say it back, that is something Amazon Polly can do. So if you think of a neurodivergent person that may be nonverbal, which some neurodivergent people are, that could be really helpful to them.

Amazon Translate as you may have seen can also help translate those words in real time.

The next example is helping to read documents. So speaking from my own personal experience, I can often really struggle with big rules of text and reading can often be quite difficult for me. So this app uses a service called Textract - I take a photo of something as I'm doing there and Textract will extract all of the text off that page for me to help me read the document or the book.

You also see that I can use the open dyslexic font to make that text more accessible if you wanted to have a read of any other examples. I've also put the QR code up there. So if you want to come back to this later, that's also ok.

Ok, so finishing up with our last theme - building diverse teams for stronger cyber defense. Anyone in cybersecurity in the audience today? We have cybersecurity teams which I'm sure we all do fabulous.

So going into some few key stats on cybersecurity diversity:

• Women make up 24% of the cybersecurity workforce.

• 9% of workers in the cybersecurity industry are people of color.

• 30% of the workforce are in the 19 to 34 age group.

• 14% of the total cybersecurity workforce have a disclosed disability - neurodivergence is included in this.

• 10% of the cybersecurity workforce identifies as LGBTQ+

• And 51% of cybersecurity professionals have military experience.

Interestingly, according to a study done by Verizon in 2022 into data breaches, 82% of data breaches involved human beings - phishing attacks, social engineering attacks, ransomware attacks often get into the system via a human.

Perry Carpenter and K. Roa said in their book, The Security Culture Playbook, "Humans are now the primary attack vector. As such, it's imperative to strengthen the human layer of defense."

Do your security teams reflect the diverse customers that you have? Can your security teams think from the perspective of that person trying to breach your system?

Diverse perspectives give diverse strengths to a team. We can strengthen our solution and product from multiple different perspectives, whether that person is neurodivergent or as part of an underrepresented community.

CJ Moses, Co-Founder of ZecOps, said "Diverse security professionals mean diverse perspectives on security, which means stronger defense." And he is absolutely correct.

So let's look at some mechanisms for building diverse cybersecurity teams:

• Over 80% of ransomware attacks come from a password breach by a human and these breaches rarely come from people in highly technical roles. So if you think of people with different life experiences and the insights they can bring and how they defended against attackers that would have come for them as a persona - that's really, really valuable.

• Over half of security professionals started outside of IT. I started outside of it. I didn't study computer science at college or university. I taught myself how to code. I actually studied veterinary science. I completely pivoted my career when I was 21. 9 years later, I'm a solutions architect.

• So when you're hiring people, don't just think about technical skills. Remember that technical skills can be learned - hire for that attitude and aptitude. Are they willing to learn? Are they passionate about the topic? Remember that diverse life experiences, different lives, different brains can bring different skill sets.

• Neurodivergent people have a way of thinking outside of the box. This is an extremely positive thing, especially in the cybersecurity industry.

I also wanted to share about a few things that we do internally at Amazon that may give you some inspiration to take back home:

• Firstly, our Amazon Aspiring Security Engineer program - we launched this in 2020 to fill the demand internally for cloud skilled security engineers. We provided internal employees with a mechanism to learn about cloud security if they weren't already security specialists and assigned them a mentor to work with them and a curriculum to eventually take them into an entry level security role. This then helped build a pipeline of cloud skilled security engineers internally within our organization and also enhance the career mobility amongst existing engineers by giving them an opportunity to mentor outside of their direct team.

• We also have the concept of Technical Field Communities at AWS. These are technological communities that allow people in technical roles to join communities that are directly related to certain subject areas. Security is one of them. I'm personally part of the Analytics one and this helps people who may not necessarily be a specialist in that area yet to collaborate with specialists and define a shared interest and perspective to different teams. And from a security perspective, this also drives proactivity and also security awareness across the organization and also gives a lot more opportunities for underrepresented groups.

Now to leave you with some key lessons learnt from the both of us:

• Firstly, I think Manisha touched on this earlier as well - limit ambiguity, be transparent in what you require and define the why of what you're asking, not just the how and the what.

• Use people first and/or identity first language, understand the needs of your customers. Remember that neurodiversity affects one in seven people. At least one of your end users are neurodivergent - are you making your systems and applications accessible to them?

• If you recognize that a system isn't accessible, identify and remove those barriers and also continue to innovate on behalf of customers with disabilities. Could that new technology that's come out recently - for example, generative AI - could that help someone with a disability or who's neurodivergent?

Finally, I just wanted to say thank you so so much for listening. We hope you found this session helpful and it's given you a few key takeaways to take that home. As Manisha said, we'll be available in the All Builders Welcome lounge and the Venetian if you want to come chat to us and we've also got time for questions also. So if you have any questions, please feel free to either come find us at the end or just ask your questions.

Now, also, just a quick reminder, please fill in the survey on your AWS Events mobile app and enjoy the rest of re:Invent. Thank you so so much!