Meta builds a new team to speed up its AI model training

In short: Meta is setting up a new internal team focused on making its AI models smarter, faster by feeding them better real-world data and feedback.

What happened

According to an internal memo reported by The Wall Street Journal, Meta is creating a new Applied AI Engineering organization. The team will be led by Maher Saba, a vice president in Meta's Reality Labs division, and will report to Chief Technology Officer Andrew Bosworth.

The new group has a specific job: build the systems that help Meta's AI models improve by learning from real-world use. Think of it like a feedback loop. The AI makes predictions, people use the product, and the team collects what worked and what did not, then sends that information back to improve the next version of the model.

Meta is also trying something unusual with how the team is organized. They want a "flat structure" with up to 50 engineers for every one manager. That is a much flatter setup than most companies, where the ratio is closer to 8 to 1. The idea is to reduce bureaucracy and move faster.

This new team will work closely with Meta's Superintelligence Lab, which was created last summer and is led by former Scale AI chief Alexandr Wang. That lab focuses on building Meta's most advanced AI models.

Why it matters

The race to build better AI models is not just about having more computing power. It is also about how quickly you can learn from mistakes and improve. Meta is betting that a dedicated, lean team focused on this feedback loop will help it pull ahead of competitors like Google and OpenAI. For regular users, this could mean smarter features in Instagram, WhatsApp, and Meta's other apps over time.

Source: The Wall Street Journal