Marketing Personalization · Real-time Analytics
Amazon Ads needed to deliver relevant product suggestions to online shoppers at the scale of tens of billions of impressions each day while serving hundreds of millions of deep learning requests per second with very low latency.
Reported impact
By combining deep product understanding and near-real-time trend analysis on AWS, Amazon Ads delivers ultralow-latency recommendations at scale and now recommends relevant products to customers in 20 markets.
Amazon Ads built an ultralow-latency recommendation stack on AWS by combining deep product understanding with near-real-time trend analysis. For product understanding, the company trains embedding representations on product titles and descriptions, then uses a hybrid caching approach with Amazon ElastiCache so popular product embeddings stay in local cache while less popular ones are fetched remotely. For trend understanding, Amazon Kinesis processes engagement data in near real time, and Amazon SQS helps prioritize near-real-time traffic over batch feature publishing so compute stays focused on fast-changing shopping behavior. This architecture lets Amazon Ads recommend relevant products at scale across 20 markets without letting latency or bandwidth costs explode.
Product descriptions and titles are embedded for product understanding, popular product embeddings are served from local Amazon ElastiCache caches while less popular products use remote cache, engagement data streams through Amazon Kinesis for near-real-time trend analysis, and Amazon SQS prioritizes near-real-time features over batch workloads.
Amazon Ads employed Amazon ElastiCache and Amazon Kinesis to process billions of impressions every day at ultralow latency.
Now, the company’s machine learning models recommend relevant products to customers in 20 markets.
To process product engagement in near real time, Amazon Ads uses Amazon Kinesis, which offers key capabilities to cost-effectively process streaming data at nearly any scale.
To scale efficiently, Amazon Ads has implemented a scalable hybrid approach using Amazon ElastiCache, an in-memory caching service supporting flexible, near-real-time use cases.
We need to deal with hundreds of millions of deep learning requests per second for online influencing within a very small amount of latency.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Alibaba Cloud Sample Oct 13, 2026 TOPSPORTS (滔搏) TOPSPORTS wanted to scale enterprise AI agents beyond isolated pilots, but every new agent risked recreating the same knowledge, data, tools, and skills in separate silos. Business-analysis, membership, and marketing… business analysis time reduced to 20 minutes AgentCore · AgentLoop · MCP tools 
Google Cloud Sample Modified · Oct 9, 2026 Central Retail Food Replaced ~2,000 monthly manual Excel reports with a single source of truth on BigQuery through Looker dashboards Central Retail Food unifies its ecosystem on Google Cloud, moving from fragmented data to conversational… Addresses 50% to 60% of customer service queries with automated chatbots built on Conversational Agents BigQuery · Gemini · Gemini platform Google Cloud Sample Modified · Oct 9, 2026 ICA Previously, however, ICA’s fragmented technical environment was holding it back. Data was siloed across legacy platforms and SQL server environments, creating high maintenance costs and preventing teams from sharing… ICA Gruppen is a leading Swedish retailer, with a portfolio covering grocery, pharmacy and banking. Serving… BigQuery · Bigtable · Gemini 
