Berlin Tech Meetup: The Future of Relational Foundation Models, Systems, and Real-World Applications

Register now:

Personalization & Recommendations

Personalization, Recommendations & Search Ranking

Move beyond collaborative filtering. Kumo learns from the full relational graph — purchases, views, returns, reviews, and session context — to deliver recommendations that reflect genuine individual preference.

Product RecsContent PersonalizationSearch RankingNext Best OfferEmail OptimizationSend-TimeNotificationsCold-Start

Book a demo and get a free trial of the full platform: research agent, fine-tune capabilities, and forward-deployed engineer support.

By submitting, you accept the Terms and Privacy Policy.

8

Prediction types

Recs to send-time

3.2x

Accuracy lift

vs. collaborative filtering

<1 hr

To production

Per use case

0

Cold-start issues

Works on new items

Loved by data scientists, ML engineers & CXOs at

Catalina Logo

8 personalization use cases

Every touchpoint. Personalized.

From product pages to push notifications — one relational foundation model powers every personalization surface.

Product recommendations

Learn from the full relational graph - purchases, views, returns, reviews - producing recommendations that reflect genuine preference.

Explore

Content personalization

Personalize content feeds, articles, and media by learning from interaction patterns across users and content items.

Explore

Search ranking

Re-rank search results using relational context - what similar users clicked, purchased, and returned.

Explore

Next best offer

Predict the most relevant offer for each customer based on purchase history, preferences, and behavior context.

Explore

Email personalization

Optimize email content, timing, and product selection per recipient using relational engagement signals.

Explore

New collection launch recs

Recommend newly launched items to users based on purchase history and relational context - even with zero interaction data on the new products.

Explore

Notification reranking

Rerank push notifications and in-app messages by predicted engagement - which message, for which user, at which moment.

Explore

Send-time optimization

Predict the optimal send time for each user based on historical open and conversion patterns across channels.

Explore

Traditional recs vs. Kumo

See how relational learning compares to collaborative filtering and content-based approaches.

Data used

Traditional

User-item interaction matrix only

With Kumo

Full relational graph across all tables

Cold-start items

Traditional

No interactions = no recs

With Kumo

Graph context from similar items & attributes

Feature engineering

Traditional

Weeks per model

With Kumo

Zero — automated from relational data

Real-time context

Traditional

Batch-only, stale features

With Kumo

Session, recency, and graph signals

Cross-domain recs

Traditional

Separate models per surface

With Kumo

One model across all surfaces

Setup time

Traditional

3–6 months

With Kumo

Under 1 hour

How It Works

Simply connect your data, start asking predictions, and get results.Want more control? Fine-tune the model for your specific use case.

Connect your data
STEP 1

Connect your data

Integrates directly with your warehouse, no additional pipeline setup.

Ask a predictive question
STEP 2

Ask a predictive question

Ask questions in plain English and let Kumo do the modeling for you.

Act on predictions
STEP 3

Act on predictions

Get clear predictions and push them instantly into your workflows.

churn_prediction.pql
PREDICT COUNT(transactions.*, 0, 90, days) = 0
FOR EACH customers.customer_id
WHERE COUNT(transactions.*, -60, 0, days) > 0

3 lines. No feature engineering. No pipeline code.

For developers

Predict in a few lines of SQL

Kumo's Predictive Query Language (PQL) replaces months of feature engineering, model training, and pipeline work with a few lines of SQL-like syntax. Describe what you want to predict — Kumo handles the rest.

PQL docs|Python SDK

Why Kumo

01Zero-Shot Foundation Models

Get accurate predictions on relational data instantly—no training or ML setup required.

Read the KumoRFM announcement
Snail
02Real-Time Predictions
03Native Data Warehouse Integration
04Fine-Tuning at Scale
05Enterprise-Grade Security
06Transparent Explainability

Built by pioneers in AI

Vanja Josifovski

Vanja Josifovski

CEO and Co-Founder

Former CTO at Airbnb and Pinterest

Jure Leskovec

Jure Leskovec

Co-Founder & Chief Scientist

Stanford Professor · Co-creator of RDL and GNN

Hema Raghavan

Hema Raghavan

Co-Founder & Head of Engineering

Former Sr. Director of Engineering at LinkedIn

Meet the full team

Loved by data scientists, ML engineers & CXOs at

Catalina Logo

Peer-reviewed

Built on world-class research

Kumo is built on 40+ peer-reviewed papers at NeurIPS, ICML, and KDD.

RFMZero-shotFine-tunedTransfer
ICML 2024

KumoRFM: A Relational Foundation Model for Predictive Analytics

K. Huang, M. Fey, J. Leskovec et al.

A foundation model for relational data — pre-trained across schemas, it delivers accurate predictions out of the box.

Read paper
ABC
NeurIPS 2024

Relational Deep Learning: Graph Representation Learning on Relational Databases

M. Fey, W. Hu, K. Huang, J. Leskovec et al.

Learning predictive models directly on relational databases, eliminating the feature engineering pipeline.

Read paper
T1T2T3T4T5+20+20+23+22+35BaselineKumo30 tasks
NeurIPS 2024 · Datasets Track

RelBench: A Benchmark for Deep Learning on Relational Databases

J. Robinson, R. Miao, K. Huang et al.

An open benchmark for evaluating relational prediction methods across 11 databases and 30 tasks.

Read paper
Explore all 40+ research publications

Deep dive

The Complete Guide to Recommendation Systems

From collaborative filtering to graph neural networks. 5 approaches compared, cold-start solutions, MAP@K benchmarks, 7-tool comparison, and 6 common mistakes.

Read the complete guide

Your user data already contains the preference signal.

See what Kumo can personalize from your existing relational database.

Request a demoTry Kumo free