Method

Multi-task learning

One network trained on several tasks with shared layers.

59 cases from 25 companies apply Multi-task learning. 24 were checked in an online A/B test; 18 report results in production.

Typical tasks: Retrieve and rank from millions of items, Recommend from what the user just did, Learn user and item embeddings, Rank search results and 13 more. Companies: Pinterest, Meta, DoorDash, Etsy, Yandex, Avito and 19 more. Often combined with: Sequential recommendation, Two-tower model, Deep & Cross Network, Knowledge distillation, Gradient boosting and 99 more.

Meta’s Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AIadtech2026-10
The Model Is Not the Modelingfintech2026-09
Beyond Two Towers: Launching the 3-Tower Engagement Co-Train Model (Part 2)adtech2026-09
Learning Richer Credit Representations with Transformers at Affirmfintech2026-09
Building a unified consumer memory for personalization at scalefood-delivery2026-06
SilverTorch: Index as Model — A New Retrieval Paradigm for Recommendation Systemssocial2026-05
Neural Network Recommendations for Smart TV: Moving from Classical MLmedia-streaming2026-05
How Transformers Increased Candidate Diversity on the Avito Home Pageclassifieds2026-05
Winner's Solution Walkthrough: 1st Place (AvitoTech ML Cup 2026)classifieds2026-05
From Clicks to Conversions: Architecting Shopping Conversion Candidate Generation at Pinterestadtech2026-04
Modernizing the Facebook Groups Search to Unlock the Power of Community Knowledgesocial2026-04
Making Ads Count: Using MMoE and Auxiliary Tasks to Better Connect Buyers & Sellersmarketplace2026-03
Friend Bubbles: Enhancing Social Discovery on Facebook Reelssocial2026-03
Recommending travel destinations to help users exploretravel2026-03
Unifying Ads Engagement Modeling Across Pinterest Surfacesadtech2026-03
GPU-Serving Two-Tower Models for Lightweight Ads Engagement Predictionadtech2026-02
Adapting the Facebook Reels RecSys AI Model Based on User Feedbacksocial2026-01
Mind the Gap: Using LLMs to bridge behavioral silos in multi-vertical recommendationsfood-delivery2025-12
Criteo Boosts Ad Performance with Latest AI Advancementsadtech2025-11
Meta’s Generative Ads Model (GEM): The Central Brain Accelerating Ads Recommendation AI Innovationadtech2025-11
DoorDash: an LLM-assisted personalization framework balancing familiarity, affordability and noveltyretail2025-10
Next-Level Personalization: How 16k+ Lifelong User Actions Supercharge Pinterest’s Recommendationssocial2025-06
On the Deep Value of Useless Email Campaignsgeneral2025-05
Journey to 1000 models: Scaling Instagram’s recommendation systemsocial2025-05
FM-Intent: Predicting User Session Intent with Hierarchical Multi-Task Learningmedia-streaming2025-05
Does High Click-through Rate Lead to High Conversion Rate?travel2025-05
Multi-gate-Mixture-of-Experts (MMoE) model architecture and knowledge distillation in Ads Engagement modeling developmentadtech2025-04
Foundation Model for Personalized Recommendationmedia-streaming2025-03
Module Relevance on Home Feedsocial2025-03
How to investigate the online vs offline performance for DNN modelsfood-delivery2024-12
Precision in Motion: Deep learning for smarter ETA predictionsfood-delivery2024-10
NEP: Notification System and Relevancesocial2024-08
Efficient visual representation learning and evaluationmarketplace2024-06
Predicting a Listing Category from Its Description and Photoclassifieds2024-05
How We Fought Spam in Reviewsmarketplace2024-05
Technology Trends in RecSysgeneral2024-05
Improving ETAs with multi-task models, deep learning, and probabilistic forecastsfood-delivery2024-03
Evolution of Ads Conversion Optimization Models at Pinterestadtech2024-01
Learning to Understand Dialogs Based on Real Datafintech2023-10
How LinkedIn Is Using Embeddings to Up Its Match Game for Job Seekerssocial2023-10
Building for Inclusivity: The Technical Blueprint of Pinterest’s Multidimensional Diversificationsocial2023-09
From Pre-trained to Fine-tuned: Nextdoor’s Path to Effective Embedding Applicationssocial2023-09
Scaling the Instagram Explore recommendations systemsocial2023-08
Deep Multi-task Learning and Real-time Personalization for Closeup Recommendationssocial2023-06
Experimenting with Machine Learning to Target In-App Messagingmedia-streaming2023-06
History of the Unified Classifier Developmentbanking2023-05
From Image Classification to Multitask Modeling: Building Etsy's Search by Image Featuremarketplace2023-05
How We Built a Multi-Task Canonical Ranker for Recommendations at Etsymarketplace2023-04
How Pinterest Leverages Realtime User Actions in Recommendation to Boost Homefeed Engagement Volumesocial2022-11
What is good, what is bad. Quality of X-ray images—healthtech2022-10