Artificial Intelligence has the potential to enhance and enable new experiences across Facebook. When you process billions of posts, comments, and images per day, you need to ensure you’re building the most efficient AI systems on the planet.

Talent and expertise in AI technologies is in short supply, especially in deep learning. Universities have increased their offerings of AI classes, but many engineers lack the background and hands-on experience to build AI systems that run in the wild.

To solve this gap we created the Facebook AI Academy, as recently written about in this Wired article. We believe that innovation relies on education. By educating our engineers in the latest AI technologies, we can deploy a broad set of deep learning experts across the company. To date, more than 40 teams at Facebook — and more than 25% of our engineers — are using AI technology to power the products and services they build. We want to grow that even further.

The AI Academy accomplishes this in two ways. The first is through a series of hands-on classes in deep learning. The curriculum, designed by myself and other leading experts in the field, brings engineers from across the company together to learn directly from the experts at Facebook AI Research (FAIR). Academy topics span deep learning basics, convolutional and recurrent neural networks as well as related topics such as reinforcement learning. The 6-class series combines academic lectures and labs where engineers build their own deep learning models.

The second part of Facebook’s AI Academy is for students seeking a more in-depth hands-on experience. To accomplish this, we created the AI Immersion Program. This program provides engineers the opportunity to join Facebook AI Research for one or two years. During this time, they work on research projects with the world’s leading experts in numerous areas, including deep learning, computer vision, natural language processing, speech, and reasoning. After graduation, the engineers leave the FAIR team to apply their knowledge to other engineering groups across Facebook, such as the Applied Machine Learning group, Newsfeed, etc. This not only spreads deeper AI expertise across Facebook, it also creates a strong community of collaborators between FAIR and the rest of the company.

The AI Academy has been an overwhelmingly (and oversubscribed) positive learning program for Facebook engineers, and has also been a tremendous learning opportunity and connector for researchers in FAIR. Research in AI has become increasingly complex; requiring efficient software platforms, large-scale compute clusters, and huge sources of data. In other words, to do leading scientific research, we need great engineers! The AI Academy has given us a much deeper connection with the amazing engineering talent at Facebook.

We also know that to expedite progress and ensure a consistent flow of new engineering talent, it is important to maintain a vibrant academic research community. In keeping with Facebook’s openness, FAIR helps the larger community by providing open source software and hardware. Check out some of our recent open source releases. The AI Academy is currently only available to Facebook engineers, but we will continue to evaluate opportunities for sharing knowledge with the broader academic community.