T-Bank's Perseus: An ML Personalization Framework Across the Entire Service Ecosystem

On July 23, 2026, T-Bank unveiled Perseus, a universal personalization framework that merges all of a user's actions across the different services in its ecosystem into a single event sequence and can raise business metrics by anywhere from 3% to 17%.

What problem does Perseus solve

Perseus closes the gap between the data of dozens of T-Bank services, from supermarket purchases to mortgage applications. Previously, each ML model learned only from the data of the service where it was deployed, so it saw the user in fragments rather than as a whole.

The problem was compounded by data heterogeneity: teams collect events independently, every service has its own schema, and new products have no behavioral history at all, the classic "cold start." According to T-Bank, this is exactly what prevented personalization from being carried over between products in the ecosystem and forced each team to build its own model from scratch.

• Author: Oleg Lashinin, head of recommender systems at T-Bank
• Problem: every service has its own event schema and its own isolated model
• New services have no user history: the "cold start"
• Solution: a single sequence of all of a user's actions
• Business metrics grow by 3% to 17% across different tasks

How the unified event stream works

Perseus stitches heterogeneous events from all services into one chronological sequence and uses it to build a holistic representation of the user. According to the developers, this approach makes the framework a flexible tool for working with ecosystem data and lets the company scale ML personalization across its products instead of retraining a separate model for each service.

The single sequence solves two problems at once: it brings events with different schemas into a common form, and it carries behavioral signal from mature services over to new ones. As a result, the model has context about the user before they have taken a single action in a given product.

"When an ecosystem has dozens of services, you want the model to understand the user as a whole, not in pieces," said

Oleg Lashinin, head of recommender systems at T-Bank, in the company's blog on Habr.

What this means for the business

The main result is metric growth even where personalization previously did not work. According to T-Bank, business metrics rose by 3% to 17% across different tasks, and the effect held even for users who had not made a single purchase in a given service.

This is the key difference from classic recommender models, which are helpless without behavioral history. Perseus instead draws on the user's actions in other ecosystem services and personalizes from the very first touch with a new product. For a company with dozens of services, this means that launching personalization in a new product no longer depends on accumulating data.

What this means

Large fintech ecosystems are moving toward a "unified behavioral profile": instead of dozens of isolated models, a single event stream on which all services train. For T-Bank, this is a way to quickly switch on personalization in new products that have no accumulated history yet, and to squeeze out up to 17% growth in business metrics.

Frequently asked questions

What is T-Bank's Perseus?

Perseus is a universal personalization framework that merges a user's actions across all services in the T-Bank ecosystem into a single event sequence. ML personalization for the company's different products is then built on this shared data.

By how much does Perseus improve metrics?

According to T-Bank, business metrics rose by 3% to 17% across different tasks, including for users without a single purchase, where ordinary models fail because there is no history.