Engineering at Meta

Home
United States
Menlo Park, CA
Engineering at Meta

Contact information, map and directions, contact form, opening hours, services, ratings, photos, videos and announcements from Engineering at Meta, Science, Technology & Engineering, Menlo Park, CA.

Engineering at Meta is a technical news resource for engineers interested in how we solve large-scale technical challenges at Meta
This page DOES NOT PROVIDE TECHNICAL SUPPORT for Meta products and services.

10/21/2026

We trained our Muse foundational models, including Muse Wonder, at Prometheus, our gigawatt-scale supercluster in New Albany, Ohio.

Across a supercluster with hundreds of thousands of GPUs, even one underperforming host can slow an entire training run. We engineered Prometheus for resilience at scale, achieving 85%+ goodput on pretraining, 90%+ on reinforcement learning and ~47K GPU hours mean time between interruptions during Muse Wonder training.

Take a look at the tech powering Prometheus: https://go.meta.me/4yXXApZ

10/20/2026

🎯 Excited to share how we scaled the 2026 generation of GEM — Meta's Generative Ads Model — the central model and compute engine behind ads across Meta.

Last year we introduced GEM, the industry's largest recommendation systems foundation model. This year, we are pushing the frontier again with breakthroughs across model architecture, cross-surface cross-domain learning at Meta’s scale, innovative intelligence transfer, and LLM-scale AI infrastructure:

Bending the scaling curve with Kunlun. GEM's new LLM-style architecture, co-designed with the latest GPU hardware, delivers 2.5x model complexity growth over earlier versions of GEM and processes 2x longer engagement history per person, while modeling the massive feature interactions unique to recommendation systems.

Expanding learning across Meta's ads ecosystem. GEM is the first model to jointly predict clicks and conversions across all surfaces — Facebook, Instagram, Reels — and across major deep-funnel objectives like off-site conversions and mobile app installs.

Doubling the effectiveness of intelligence transfer with LoopFM. This new frontier post-training technique passes GEM's internal representations and historical memory to the models that serve ads as a sequential context window, at no additional compute cost or latency. Elastic pre-computation (Solaris) further harvests gains from upstream models.

Scaling frontier training efficiently. Through deep model and infrastructure co-design, we increased total training FLOPs by 4x while doubling end-to-end training efficiency to 20–25% MFU over the past 12 months, across thousands of latest-generation GPUs.
Recommendation systems don't scale the way language models do — heterogeneous features, shifting distributions, and no option to serve a frontier model in a live request. Making GEM scale like LLM took rethinking the scaling architecture, the post-training and the training stack together. Very proud of this team.

📈 The payoff: for people, ads that better match what they're actually looking for. For advertisers, an 18.9% lift in conversions and 12% in click-through rate across our ad surfaces and objectives over the past year — with 5.8%, 4.1% and 1.6% conversion uplift on Facebook, Instagram and off-site conversion for targeted users.

🔗 Read the full engineering story: https://engineering.fb.com/2026/10/13/ml-applications/generative-ads-model-gem-scaling-frontier-ads-ai/

10/06/2026

We’re open-sourcing Rebalancer, a generic, high-performance library designed to solve complex, large-scale resource assignment problems.

Used continuously in production at Meta for over a decade, Rebalancer manages infrastructure allocation challenges by decoupling how an assignment problem is specified from how it is solved.

Traditional optimization methods often force a choice between rigid solver-specific formulations or ad hoc local search heuristics. Rebalancer bridges this gap with a domain-specific expression graph API and intuitive modeling constructs (Dimensions, Partitions, Scopes, and Utilization).

Key Technical Capabilities & Performance Highlights:

→ Flexible Solving Architecture: Evaluates local search heuristics for large-scale operations or generates Mixed Integer Programs (MIP) solvable via HiGHS, Gurobi, or FICO Xpress.
→ Proven Hyperscale Scalability: P99 solve times of ~12 seconds on 265k objects across 3.2k bins. For extreme workloads (>1M objects, 5k bins), average solve times reach 171 seconds.
→ Rebalancer Explorer: Includes a Dockerized UI for rapid debugging, helping engineers understand constraint bounds, evaluate relaxations, and inspect local search decisions.

Whether applied to datacenter task placement, energy grid logistics, or large-scale compute cluster balancing, Rebalancer provides a unified framework for resource allocation.

🔗 Read our engineering blog: https://engineering.fb.com/2026/09/21/open-source/rebalancer-generic-high-performance-library-assignment-problems/

🔗 Explore the GitHub repo: https://github.com/facebook/rebalancer

🔗 Read our OSDI '24 research paper:https://www.usenix.org/system/files/osdi24-kumar.pdf

09/24/2026

We want AI glasses to be about more than just hands-free calling and messaging, but advanced features require more compute than can fit into a pair of glasses.
Enter Private Processing.

We first introduced Private Processing last year in WhatsApp and the Meta AI app to keep your chats with Meta AI completely private. Now, as we announced at , we’re expanding Private Processing to bring these same privacy benefits to our AI glasses.

Read the technical deep dive on our approach: https://engineering.fb.com/2026/09/23/security/private-processing-meta-ai-glasses/

09/18/2026

To help Meta subject matter experts focus more on work where their knowledge matters most, we’ve built an AI agent that acts as a secondary expert for a given domain. This system makes deep specialist knowledge readily available and preserved for anyone in an organization to access, share, and build upon.

Two unique features of this approach:

1️⃣ A structured, auditable knowledge architecture that separates what the agent knows from how it reasons. Reasoning is handled separately by composable "recipes" that prescribe analytical workflows without containing domain facts.

2️⃣ A self-improvement loop that turns expert feedback into verified updates without retraining. It treats knowledge maintenance as a compilation problem: diagnose feedback to its root cause, compile minimal verified edits, validate against a regression suite, and route to a human expert for review. Every landed fix is added back to the regression suite, so future changes must preserve corrected behavior.

Human experts stay in control throughout. The agent surfaces its intermediate reasoning at defined checkpoints, where experts can confirm, correct, or redirect. Every correction feeds the improvement loop.

Read the full engineering deep dive: https://engineering.fb.com/2026/09/02/ml-applications/organizational-second-brain-ai-learns-from-experts/

09/14/2026

At Meta’s scale, managing over 1 million clients and 500k+ database hosts, direct database access stops being efficient and becomes a massive reliability risk. Our latest engineering deep dive breaks down ZGateway, the stateless proxy tier now unifying traffic for ZippyDB (>1B ops/sec).

By decoupling the client and server fleets, ZGateway collapses a massive many-to-many TLS connection mesh. The results: Per-host connection counts dropped by 97–98%, and total persistent connections fell by 19x, all for just ~6% compute overhead.

Read the full deep dive here: https://engineering.fb.com/2026/09/03/core-infra/zgateway-proxy-zippydb-meta/

09/01/2026

We’re releasing MetaRoCE through the Open Compute Project (OCP), making our complete specification, reference software implementation, and compliance test suite freely available to the broader tech industry.

Training and serving frontier AI models requires ultra-fast, reliable networks that transfer data between GPUs without wasting valuable compute cycles. To solve this at scale, we designed MetaRoCE, a clean-sheet RDMA transport protocol purpose-built specifically for AI workloads on commodity Ethernet.

By contributing this to OCP, we are enabling hardware vendors, network operators, and developers to adopt, implement, and build the future of AI infrastructure together.

Watch the full technical deep dive from the At Scale Networking Conference: https://engineering.fb.com/2026/08/24/networking-traffic/metaroce-rdma-transport-ai-ethernet/

08/31/2026

When embedding tables make up >99% of model parameters, GPUs lose up to 20% of compute throughput just managing AllReduce and AllToAll collectives.

MTIA 300, the first of Meta’s family of in-house training and inference accelerators optimized for training ranking and recommendation models, fixes this by co-designing hardware and networking from scratch:

1️⃣ On-package NICs: 12x custom 800 Gbps RDMA NICs (1.2 TB/s I/O) bypass the host PCIe bus entirely.

2️⃣ Offloaded ex*****on: 16 dedicated RISC-V Message Engines handle network ops, keeping compute degradation under 0.5%.

3️⃣ Compiled HCCL: Integrates directly with torch.compile and PyTorch c10d for host-free ex*****on.

The result: 3.9x faster communication time on a 150B-parameter production recommendation model.

See the technical deep-dive from the At Scale: Networking Conference: https://engineering.fb.com/2026/08/24/networking-traffic/mtia-300-meta-training-chip-built-in-nics/

08/06/2026

Meta has developed and deployed a new multi-layer sequence learning architecture for ads recommendation systems.

This architecture is built for production at scale as a resource efficient, multi-stage system that boasts high upstream-to-online transfer ratios.

📈 Our latest research paper, “Scaling Laws for Multi-Stage Sequence Modeling in Large-Scale Ads Recommendation,” contains a comprehensive scaling law study — analyzing model width, depth, information density, and sequence length — and how that drives ad performance growth.

👉 Read the full paper to learn more! https://engineering.fb.com/2026/08/05/ml-applications/from-user-sequences-to-scaling-laws-a-multi-stage-architecture-for-metas-ads-ranking/

07/23/2026

At Meta's scale, a few milliseconds of kernel scheduling latency can impact millions of ad decisions per second.

When we upgraded our Linux kernel across the ads serving fleet, the new default scheduler introduced a latency regression that reduced the number of ads we could rank in time. We needed a solution.

Enter sched_ext: an open-source, BPF-based framework we co-developed that lets engineers write and deploy custom scheduling policies without touching the kernel itself.

The results speak for themselves:�
📉 28% reduction in P99 tail latency�
⚡ 3.28 megawatts in power savings�
📈 1.1% increase in ads ranked

Now, sched_ext is upstream in Linux v6.12. That means any team — at a hyperscaler, cloud provider, or embedded systems company — can build workload-specific scheduling without forking the kernel.

Read the full deep dive: https://engineering.fb.com/2026/07/13/ml-applications/modernizing-the-meta-ads-service-with-an-open-source-kernel-scheduler/

Address

Menlo Park, CA
94025

Website

https://engineering.fb.com/

Alerts

Be the first to know and let us send you an email when Engineering at Meta posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Engineering at Meta:

Want your business to be the top-listed engineering company in Menlo Park?