Meta Andromeda
Meta Andromeda is a personalized ads retrieval engine built in partnership with NVIDIA that applies machine learning to the earliest stage of Meta's ad delivery pipeline.
Last updated Nov 19, 2026 by ATDb automated enrichment
AI-drafted from public sources10 sources cited
At a glance
About
Meta's proprietary first-stage ads retrieval engine powering personalized ad delivery across Facebook and Instagram at massive scale
Meta Andromeda is a personalized ads retrieval engine built in partnership with NVIDIA that applies machine learning to the earliest stage of Meta's ad delivery pipeline. It operates as the first step in a multi-stage recommendation system, narrowing tens of millions of ad candidates down to a few thousand relevant options before more sophisticated ranking models determine the final ads shown to a user. The system uses a deep neural network with increased compute complexity running on the NVIDIA Grace Hopper Superchip, and its deployment across Instagram and Facebook has achieved a six-percent recall improvement in retrieval and an eight-percent improvement in ads quality on selected segments. Andromeda supports Meta's Advantage+ suite, which automates budget allocation, audience targeting, and bid adjustments, with advertisers who activated its AI-driven targeting features reporting a 22-percent increase in return on ad spend. The engine is designed to scale with the growing volume of ad creatives generated through generative AI tools, using a hierarchical index built on deep neural networks to handle that expansion efficiently. Source 1: fb.com: Meta Andromeda: Supercharging Advantage+ automation ...
Business model
Internal Product
Target market
Enterprise
What they offer
Andromeda Retrieval Engine
First-stage ML-based ads retrieval system that narrows tens of millions of ad candidates to a few thousand relevant options for downstream ranking models.
Hierarchical Index
A deep neural network-based hierarchical index designed to efficiently scale with the growing volume of ad creatives, including those generated by generative AI tools.
NVIDIA Grace Hopper Integration
Deep neural network with increased compute complexity running on the NVIDIA Grace Hopper Superchip to power high-performance retrieval.
Features, use cases and customers
7 key features · 4 use cases · 2 customer segments
Key features
Use cases
Customer segments
Tech & specs
4 technologies · 3 security and compliance items · 1 deployment option · 1 API
Technology stack
Security & compliance
Deployment
API
No