YooMoney builds a RAG platform and "Eugene," an AI assistant for searching its corporate wiki

Payments company YooMoney has explained on Habr how it built its own RAG (Retrieval-Augmented Generation) platform and an AI assistant called Eugene. The assistant replaced standard search on the corporate wiki and understands what an employee's query means, rather than just matching keywords.

Why standard search stopped working

Over time, companies pile up thousands of pages of internal documentation, and finding the right information gets harder and harder. According to YooMoney, standard search on a corporate wiki can no longer cope with that volume. Employees waste time clicking through pages, and instead of finding answers on their own, they ask colleagues the same questions over and over.

• Company: YooMoney, a Russian payments service
• Assistant name: Eugene, built on RAG
• Author: Ilya, head of YooMoney's development department
• The team's key decision: rejecting off-the-shelf RAG solutions in favor of an in-house platform
• Target audience: architects, AI engineers, and specialists deploying enterprise LLM solutions

How the Eugene AI assistant works

Eugene combines the company's internal knowledge with the capabilities of large language models. RAG stands for Retrieval-Augmented Generation. First, the system pulls documentation snippets relevant to the question from the knowledge base. Then the language model writes an answer based on those snippets, rather than relying only on the data it was originally trained on. This lets the system answer based on what a query means, not on literal word matches, as classic full-text wiki search does.

According to the YooMoney team, it was the arrival of large language models as practical working tools that made it possible to combine the company's accumulated knowledge with LLM capabilities. That is how Eugene came about: an assistant that can understand what an employee is actually asking, not just the keywords in the question.

Why YooMoney passed on off-the-shelf solutions

Ilya, head of YooMoney's development department, promises to explain why the team turned down the RAG solutions available on the market and chose to build its own corporate knowledge platform instead. He writes primarily for architects and AI engineers facing a similar choice when rolling out LLMs inside their companies: whether to plug in an existing product or build a platform around their own corporate data.

YooMoney pitches the piece as a practical breakdown of exactly that decision: why building in-house turned out to be preferable to buying or connecting a third-party RAG service for searching internal company documentation.

"My name is Ilya, and I head the development department at YooMoney.

I'll explain why we turned down off-the-shelf solutions and how we built a RAG platform for corporate knowledge," the author writes on Habr.

What it means

YooMoney illustrates a path typical for corporations in 2026: moving from scattered wiki documentation to an internal RAG-based AI assistant. RAG is becoming the standard for knowledge search at companies with large volumes of internal documentation, and YooMoney's write-up is aimed primarily at people deploying similar solutions themselves.