Mawdoo3’s artificial intelligence (AI) initiatives run on AWS, giving site visitors new ways of accessing information and a faster experience. Its online encyclopedia, launched in 2012, is the world’s largest Arabic-language content platform. It delivers in-depth content articles that are quality-checked by experts in various disciplines. Mawdoo3 relies on technologies such as AWS Elastic Beanstalk and Amazon API Gateway to develop Arabic voice-to-text services and other AI functionality.
A leading startup in the Middle East and North Africa, Mawdoo3 provides the largest Arabic-language content platform in the world. Its 84 million site visits and 42 million unique visitors a month make the online encyclopedia the most visited site in the region. The company employs 80 full-time staff and hundreds of freelancers, who create the site’s original content. There are currently more than 140,000 articles live on the site. The portal, which launched in 2012, is used by Arabic-speaking people around the world to access information in their native language.
Mawdoo3’s cofounders, Mohammad Jaber and Rami al-Qawasmi, aim to not only expand the content repository and user base, but also create new ways of accessing knowledge through artificial intelligence (AI). “We got to the point where our existing cloud infrastructure didn’t have the performance we required to realize our growth and innovation ambitions,” says Chief AI Officer Hussein Al-Natsheh. “It also didn’t have the breadth of services to allow us to develop AI functionality such as voice-to-text services. We believe AI is key to building intelligent search and retrieval that enhances content accessibility for our users.”
The opportunity to join the Amazon Web Services (AWS) Activate for Startups program inspired Mawdoo3 to move its infrastructure to the AWS Cloud in 2016. It is all-in on AWS and runs its mobile and web presence as well as its AI projects using AWS technologies.
One of Mawdoo3’s new projects is a mobile app called Gharred, which allows users to post tweets and other social media via voice. It uses Amazon CloudWatch to keep track of the generated text along with users’ voice files, helping to enhance its AI models
In another project—a service that will provide APIs to developers—Al-Natsheh and his team are using Amazon Elastic Container Registry and the Amazon API Gateway. He says, “Because deep learning is highly compute-intensive, knowing we have the dynamic capacity to power our AI development work is awesome. AWS gives us the freedom to innovate. We have a well-integrated infrastructure and the flexibility to use open-source frameworks such as TensorFlow on AWS.”
It’s not just AI functionality that’s helping Mawdoo3 make content more accessible. The startup’s ability to scale up and down according to demand is also crucial. “We rely heavily on AWS Elastic Beanstalk,” says Al-Natsheh. “Traffic to our website and mobile app isn’t predictable so AWS Auto Scaling helps us dynamically provision compute resources to help ensure site visitors have a great experience.”
Jaber and al-Qawasmi believe migrating to AWS has provided Mawdoo3 with significant business value. “We’ve doubled performance and cut web and mobile response times in half, so visitors to Mawdoo3 get a better experience,” says Jaber. “On an infrastructure level, we’re 22 times more efficient. That’s taking into account everything from monitoring to database management and ensuring system security. It’s all faster in AWS. Working in the AWS Cloud also costs about 88 percent less than our previous cloud infrastructure.”
Most significantly, though, AWS has given the company the time, resources, and technology to focus on its AI initiatives. “Without AWS, we wouldn’t have been able to grow our business through AI,” says Al-Natsheh. “AWS enables us to develop AI services that are unique. For example, by employing natural-language understanding techniques, we can enable voice search that works across a range of Arabic dialects. This means visitors to Mawdoo3 can ask questions in their own dialect rather than having to think of the equivalent terminology in standard Arabic. It’s all about helping users access the right content faster.”
Learn more about deep learning.