AWS Government, Education, & Nonprofits Blog

Personalizing studying with machine learning: Course Hero’s approach

Different students learn in different ways. While many traditional classrooms continue to rely on a one-size-fits-all approach, Course Hero delivers personalized learning to every student through its innovative, machine learning (ML)-powered online platform.

Operating under the slogan “Master your Classes,” Course Hero was founded with the vision of a world where every student graduates confident and prepared. The platform provides students access to study materials, including study guides, class notes, and practice problems. The service also includes on-demand access to subject matter expert tutors, available to help students 24/7.

“I always tell my students that I am adapting to them, rather than asking them to adapt to me in a one-size-fits-all teaching approach,” says Adrian Isaza, an Instructor in the Alternative Medicine Program at Everglades University and a contributor to the Course Hero platform. “We all want to see more students walking up to the podium on graduation day, proud of what they’ve accomplished, and ready to build great lives for themselves. We aim to help them get there.”

The dreams on the website are lofty, but Saurabh Khanwalkar, the Director of Machine Learning, Search Sciences at Course Hero, tells us, “We’re a passionate team, and we all care deeply about building the optimal learning destination for every college course.”

To get there, the team uses Amazon Web Services (AWS). Particularly important are the AWS services that help validate new course material documents. With over 25 million course-specific materials in their library right now – and counting – it would be difficult for a human team to keep up with the verification of all these new materials.

Rather than delay students’ access to new materials while humans conduct reviews, the team strengthened its cloud capabilities, growing the Course Hero Engineering team six-fold since 2014. Now, new items are added to Course Hero’s library after being checked by Amazon EMR and Amazon SageMaker Inference Pipelines. This flow ensures that the documents are not fraudulent, in violation of their honor code violations, or otherwise unusable – while saving its human employees for higher-order work.

Another area where AWS services are crucial is in enabling content discovery. Khanwalkar explains, “Students’ and educators’ searches are interesting because their intents differ. Sometimes we see a specific term (like ‘entropy’) and other times, it’s a whole subject area (like ‘economics’). We analyze searches using machine learning on AWS and we work hard on delivering relevant results no matter what.”

In order to search through the content, it must first be labeled, and Amazon SageMaker Ground Truth does that quickly and accurately. The search experiences are then built on top of Amazon Elasticsearch Service, with custom implementations written through its API extension plugin. Because there is often specific vocabulary for a subject and capturing it just right is critical to delivering the right results, the flexibility of these custom implementations is valuable. Delivering the right results also requires ML to predict what else would be useful study material for that student beyond what their search directly includes.

By personalizing the experience in this way, Course Hero provides each user with relevant content. Where a class lecture might fall short, Course Hero fills in the gaps and helps students gain mastery over particular concepts. And when a student finds an area particularly interesting, Course Hero has the resources for them to go a level deeper. Also, when an educator is seeking content for their own learning or for their classroom, Course Hero is there for them, too.

Underneath it all, AWS is there to power every aspect of the platform. “We’re a full AWS shop,” Khanwalkar smiles.