AI and the data platform: connecting AI to all an organization’s data (#Data)
On October 12, Nikolay Golov returns to Research Insights Made Simple for a livestream continuing our conversation about data platforms. This time, we will explore how to bring operational databases, analytics, and AI agents together, and whether that can eliminate the familiar zoo of integrations.
Nikolay Golov is the product director of a data platform available in the Russian market. Previously, he led data platforms at Avito and ManyChat, and he teaches databases and storage systems at Harbour.Space. We will discuss both the technologies and the practical work of building platforms that someone must then maintain.
The diagram looks appealing: retain PostgreSQL for transactions, send changes from other systems through CDC into object storage, add a Lakehouse, and connect agents through MCP. The data is collected, tools are available, and you can ask questions. There is still quite a lot of engineering between “we put everything in S3” and “the agent correctly understood the business” :)
We will look at how Databricks, Snowflake, ClickHouse, Microsoft Fabric, Google, AWS, and Oracle approach the problem. Some unify storage, some use replication and shared queries, and others build everything inside a single DBMS. Similar claims about a unified platform conceal different architectures and restrictions.
We will discuss:
- The boundaries between OLTP, analytics, and an agent backend;
- What “one copy of the data” and zero-ETL mean in practice;
- When CDC and a Lakehouse simplify life, and when they add another system to monitor;
- The role PostgreSQL, DuckLake, and S3 might play in a platform of your own;
- What an agent needs beyond table access: metric definitions, permissions, context, and a way to verify results.
I want to understand which complexity these approaches actually remove and which merely moves elsewhere. The episode will be useful for architects, data engineers, and managers choosing a platform or connecting AI to existing data.
Join us on October 12 and bring your questions and hard-earned lessons.
#Architecture #DistributedSystems #AI #AI4SDLC #Management #Leadership #Engineering