Engineering articles from Pinterest
AI summaries and key learnings from Pinterest engineering teams.
Unified Context-Intent Embeddings for Scalable Text-to-SQL
The article outlines Pinterest's evolution from basic Text-to-SQL systems to a sophisticated Analytics Agent that leverages unified context-intent embeddings for enhanced query understanding and SQL...
Unifying Ads Engagement Modeling Across Pinterest Surfaces
The article presents a comprehensive approach to unify ads engagement modeling across different surfaces at Pinterest, addressing the challenges posed by previously independent models. It outlines...
Bridging the Gap: Diagnosing Online–Offline Discrepancy in Pinterest’s L1 Conversion Models
The article discusses the challenges faced by Pinterest in reconciling offline and online performance metrics of their L1 conversion models. It highlights the discrepancies observed between strong...
Piqama: Pinterest Quota Management Ecosystem
The article introduces Piqama, Pinterest's comprehensive quota management ecosystem designed to oversee resource quotas across various systems. It outlines the architecture of Piqama, emphasizing its...
Drastically Reducing Out-of-Memory Errors in Apache Spark at Pinterest
This article details Pinterest's approach to significantly reduce out-of-memory (OOM) errors in their Apache Spark applications through a feature called Auto Memory Retries. By automatically...
GPU-Serving Two-Tower Models for Lightweight Ads Engagement Prediction
The article presents a significant advancement in Pinterest's ads recommendation system through the introduction of a GPU-serving two-tower model for lightweight ranking. This model architecture...
Next Generation DB Ingestion at Pinterest
The article outlines Pinterest's transition from a legacy batch-oriented database ingestion system to a modern, real-time ingestion framework utilizing Change Data Capture (CDC) technologies. The new...
Beyond Two Towers: Re-architecting the Serving Stack for Next-Gen Ads Lightweight Ranking Models…
The article details the transition from a traditional Two-Tower architecture to a more complex GPU-based model inference system for next-generation ad ranking. It highlights the limitations of the...
Ads Candidate Generation using Behavioral Sequence Modeling
The article outlines Pinterest's innovative approach to enhancing ad candidate generation through behavioral sequence modeling. By leveraging a transformer-based model, the team predicts user...
PinLanding: Turn Billions of Products into Instant Shopping Collections with Multimodal AI
The article presents PinLanding, an innovative pipeline designed to generate shopping collections from vast product catalogs using multimodal AI techniques. It emphasizes the transition from...
LLM-Powered Relevance Assessment for Pinterest Search
The article presents a methodology employed by Pinterest Search to enhance search relevance assessment using fine-tuned large language models (LLMs). It addresses the challenges of traditional human...
How Pinterest Built a Real‑Time Radar for Violative Content using AI
Pinterest has developed a real-time radar system to measure the prevalence of policy-violating content using AI. This system addresses historical challenges in content moderation by leveraging...
Improving Quality of Recommended Content through Pinner Surveys
The article discusses Pinterest's innovative approach to enhancing the quality of recommended content through user feedback collected via surveys. By leveraging machine learning models trained on...
On the (re)-prioritization of open-source AI
The article outlines Pinterest's strategic shift towards utilizing open-source AI models, emphasizing their cost-effectiveness and performance advantages over proprietary models. It discusses the...
Autonomous Observability at Pinterest (Part 1 of 2)
The article outlines Pinterest's journey towards enhancing its observability tools by integrating AI-driven solutions and the Model Context Protocol (MCP). It highlights the challenges posed by...
Next-Level Personalization: How 16k+ Lifelong User Actions Supercharge Pinterest’s Recommendations
The article presents an in-depth exploration of Pinterest's TransActV2 model, which enhances user personalization by leveraging a comprehensive history of user actions. By integrating a Next Action...
Unlocking Efficient Ad Retrieval: Offline Approximate Nearest Neighbors in Pinterest Ads
The article explores the implementation of Offline Approximate Nearest Neighbors (ANN) for ad retrieval at Pinterest, highlighting its advantages over Online ANN in terms of cost efficiency and...
Scaling Pinterest ML Infrastructure with Ray: From Training to End-to-End ML Pipelines
The article outlines how Pinterest has expanded the capabilities of Ray beyond traditional training and inference tasks to create a comprehensive machine learning infrastructure. It details the...
Next Gen Data Processing at Massive Scale At Pinterest With Moka (Part 1 of 2)
The article outlines Pinterest's transition from a Hadoop-based data processing platform to a Kubernetes-based architecture, specifically leveraging Spark on AWS Elastic Kubernetes Service (EKS). It...
Debugging the One-in-a-Million Failure: Migrating Pinterest’s Search Infrastructure to Kubernetes
The article chronicles the migration of Pinterest’s search infrastructure, Manas, to Kubernetes, highlighting a significant performance issue where one in a million search requests experienced...