Engineering at Meta

Our Ranking Engineer Agent (REA) autonomously executes key steps across the end-to-end machine learning lifecycle for ads ranking models. REA reduces the need for manual intervention, managing asynchronous workflows spanning days to weeks through a hibernate-and-wake mechanism, with human oversight at key strategic decision points. Read this post that covers REA’s ML experimentation capabilities: autonomously generating hypotheses, launching training jobs, debugging failures, and iterating on results: go.meta.me/3f91d5

@Ciena @Meta We're thrilled to share this milestone with Ciena as we continue to strengthen capacity and resilience across our global subsea infrastructure!

Ciena and @Meta achieve world record 800 Gb/s across 16,608 km on transpacific submarine cable. Learn more in this blog: web.ciena.com/1Q1oG1

We’re excited to announce the formation of the Optical Compute Interconnect Multi-Source Agreement group, alongside founding members @AMD, @Broadcom, @Microsoft, @nvidia, and @OpenAI. This consortium marks a critical shift toward an open, hyperscaler-driven ecosystem for the optical interconnect needs of modern AI infrastructure. Learn more: businesswire.com/news/home/2026…

Custom silicon is critical to scaling next-gen AI. We’re detailing the evolution of the Meta Training and Inference Accelerator (MTIA), our homegrown silicon family designed to power the next era of AI experiences. Traditional chip cycles span years, but model architectures change in months. To close this gap, we’ve accelerated MTIA development to release four generations in just two years. See our roadmap and tech specs here: go.meta.me/16336d

It’s our priority to keep users safe while using Messenger. In addition to end-to-end encryption that ensures direct messages and calls are protected, we also implement Advanced Browsing Protection (ABP), which protects the privacy of the links clicked on within chats while still warning people about malicious links. Read the technical deep dive on ABP here: engineering.fb.com/2026/03/09/sec…

🚀 RCCLX is now open source! Built for AMD platforms, this enhanced version of RCCL is fully integrated with Torchcomms to unlock massive speedups for your AI workloads. Featuring Direct Data Access (DDA) and Low-precision (LP) Collectives, we're making GPU communication faster than ever. Learn more: go.meta.me/3ecc91

Meta 🤝 AMD Today we’re announcing a multi-year agreement with @AMD to integrate their latest Instinct GPUs into our global infrastructure. With approximately 6GW of planned data center capacity dedicated to this deployment, we’re scaling our compute capacity to accelerate the development of cutting-edge AI models and deliver personal superintelligence to billions around the world. Learn more: go.meta.me/220f12

Today we’re breaking ground on a new state-of-the-art data center in Lebanon, Indiana – one of our largest infrastructure investments to date. This data center is designed to deliver 1GW of capacity to support both AI workloads and our core products. ➡️ Learn more: about.fb.com/news/2026/02/m…

How do you catch critical bugs at the scale of 100M+ lines of code? 🛠️ In our new research paper, we explore Just-in-Time (JIT) catching test generation. Unlike traditional hardening tests that pass at generation, JIT tests are designed to fail, surfacing regressions before code lands. This research demonstrates that JIT catching is a vital tool for maintaining stability in high-velocity, large-scale development environments. Read the full paper: go.meta.me/bf4344

The convergence of sequence modeling and recommendation systems has opened new frontiers in artificial intelligence. 🚀 Our new multi-layer sequence learning architecture demonstrates LLM-like scaling law, while delivering performance in latency-sensitive ads ranking models. This new architecture is deployed for user representation within Meta’s ads recommendation as a resource efficient, multi-stage system that boasts high upstream-to-online transfer ratios. 👉 Read our research paper to learn more: arxiv.org/abs/2601.20083

Introducing the Improved Meta Lattice: Redefining Industry-Scale Ads Recommendations Meta Lattice tackles the scalability challenges of industrial recommendation models by combining cross-domain knowledge sharing, data consolidation, model unification, distillation, and system optimizations. Lattice’s model space redesign enables: - Cross-domain knowledge sharing - Data consolidation - Unified foundational learning - Hierarchical distillation for fast inference At Meta, Lattice has helped us achieve: ✅10% increase in revenue-driving metrics ✅ +6% conversion rate ✅ 20% capacity savings ➡️Read the research paper and learn more about Meta Lattice: arxiv.org/abs/2512.09200

Today we’re announcing an up to $6 billion multi-year agreement with @Corning to supply fiber optic cables for our U.S. data centers. Fiber optics are essential to our AI infrastructure, supporting everything from wearable technology like Ray-Ban Meta AI glasses to the apps that connect billions of people and businesses worldwide. Learn more here: about.fb.com/news/2026/01/m…

We’ve advanced Facebook Reels recommendations by directly modeling user feedback, not just engagement metrics. Our new User True Interest Survey model uses large-scale surveys to better match true interests, boosting accuracy, engagement, and satisfaction. Large-scale A/B tests with 10M+ users show increased engagement (+5.2%), higher satisfaction (+5.4% in high survey ratings), and reduced integrity violations (-0.34%). Read a deep dive on our approach: go.meta.me/bce870

Supporting nuclear energy development strengthens our country’s energy infrastructure and helps create a more reliable electric grid, which are key to powering the economy and securing America’s energy independence and global leadership in AI.

Today we’re announcing landmark nuclear energy agreements that will unlock up to 6.6 GW of clean, reliable power to support the next generation of AI infrastructure and data centers. Learn more here: about.fb.com/news/2026/01/m…

We’re taking a deep dive into the engineering behind the Meta Ray-Ban Display, our most advanced AI glasses to date. On the latest Meta Tech Podcast, Kenan and Emanuel from the Wearables org join Pascal Hartig to discuss all the unique challenges of designing game-changing wearable technology, from the unique display technology to emerging UI patterns for display glasses. Listen here: engineering.fb.com/2025/12/17/vir…

We’ve released Ax 1.0: an open-source platform for adaptive experimentation at scale. Ax uses ML to automate complex, resource-intensive experiments, enabling efficient optimization for AI, infrastructure, and hardware. Learn more about Ax’s architecture, methodology, and real-world impact: engineering.fb.com/2025/11/18/ope…

Project Waterworth is redefining the scale and complexity of subsea infrastructure, connecting five continents with over 50,000 km of high-capacity cable. From mapping shipwrecks to trenching through volcanic zones, Project Waterworth is an engineering milestone. Meta subsea cable engineer Andy Palmer-Felgate shares how the team is tackling the challenge to support the next generation of AI-driven data centers and global connectivity. Read more: pcmag.com/news/inside-me…

Instagram for iOS now supports Dolby Vision and ambient viewing environment (amve) metadata, enhancing HDR video playback. Learn how we enabled end-to-end Dolby Vision support and improved video fidelity for iPhone users. 👇 engineering.fb.com/2025/11/17/ios…

StyleX is Meta’s open-source styling system for large-scale apps, combining CSS-in-JS ergonomics with static CSS performance. Now standard across Meta and adopted by Figma, Snowflake, and more. Learn how StyleX compiles expressive, type-safe styles to atomic CSS: engineering.fb.com/2025/11/11/web…