Meta creates new AI unit in Reality Labs to accelerate model development

Meta Establishes Applied AI Engineering Team to Boost Model Development

Meta has created a new AI unit within its Reality Labs division, marking another strategic shift in the company's approach to artificial intelligence development 1

1

. The applied AI engineering team will operate under the leadership of Maher Saba, a vice president in Meta's Reality Labs division, and report to Chief Technology Officer Andrew Bosworth 2

2

. This organizational change aims to accelerate model development by focusing on real-world data collection, feedback mechanisms, and evaluation processes that transform strong models into industry leaders.

Focus on Real-World Data and Reinforcement Learning

The new organization includes two core teams: one responsible for building interfaces and tooling, and another focused on executing tasks, generating data, and providing evaluations for modeling teams 1

1

. "Building great models isn't just about researchers and compute; it requires real-world data, feedback and evals," Saba wrote in an internal memo. "This creates the flywheel that turns a strong model into a leading one." The company has observed significant gains from reinforcement learning and post-training techniques, prompting this intensified focus on superintelligence initiatives. The teams will adopt an ultra-flat structure with ratios of up to 50 employees reporting to one manager 2

2

.

Strategic Positioning Within Reality Labs Division

By placing this AI unit within the Reality Labs division, Meta connects its model development efforts directly with its metaverse and AI glasses projects 2

2

. This positioning suggests the company intends to leverage Meta's AI models for practical applications in AI wearables and augmented reality products. The move follows Meta's creation of its Superintelligence Lab during summer 2025, which works on foundation models, product development, Fundamental AI Research (FAIR), and next-generation models 1

1

.

Aggressive AI Investment Strategy and Organizational Changes

This restructuring represents the fourth such reconfiguration of Meta's AI efforts in six months, occurring in August 1

1

. The company paused hiring for AI professionals after a massive recruitment drive that saw CEO Mark Zuckerberg personally involved, offering nine-figure compensation packages in some cases. Despite workforce adjustments, including eliminating 10% of jobs in Reality Labs in January to reduce focus on virtual reality products, Meta continues committing billions to AI infrastructure. The company belongs to a small group of technology giants expected to spend more than $500 billion combined on capital expenditures in 2026, largely driven by investments in data centers, chips, and AI infrastructure 1

1

.

Content Licensing Deal Fuels Training Data Strategy

Meta has secured a content licensing deal with News Corp that will run for at least three years, paying up to $50 million annually to use media content from the United States and United Kingdom 1

1

. The company will use this content to train its AI models and provide current information about events to users of its AI products. This partnership, first discussed in September, addresses the critical need for high-quality, up-to-date training data that Saba emphasized in his memo about building leading models through real-world information and continuous evaluation.