Vadim Opolski
My name is Vadim Opolski, and I’m AI guild lead at DXC Technology. It’s not a management position, and like you, I write code every day. I joined the company in 2019 as a Data Engineer and have worked on projects for clients like ATT, DirectTV, and Deutsche Bank.
I’ve designed data processing systems using Hadoop, Spark, Spark Streaming, Flink, and Akka.
Prior to Luxoft, I worked on a chat startup, and after 3 months, our team of 5 people launched the chat using Rasa, Scala, Akka, and Cassandra. It grew to over 10,000 users, and Facebook invested almost $1.5 million in the project.
I have also consulted companies such as Fortum, Mercedes-Benz, Toyota, GridGain, HSBC Bank, and BMW on big data projects.
I teach technologies such as Spark, Flink, Akka and Scala to students. I’ve also been a speaker on several interbnational conferences, including Highload 2022 and 2023, and Data Science Conference.
The Talking Lakehouse: LLM-Powered Data Platform with Trino, Spark and Superset
When people who don’t speak SQL need information kept in a relational database, they ask an analyst or a developer to retrieve the information and perhaps create a report with visualizations. This creates extra work and extra communication, and modern AI-based technologies strive to eliminate the overhead. I want to tell you about taking open-source components and building a system with a chat-bot interface that takes human language requests, transforms them into SQL and returns back to the user with answers and visualizations. I will describe a multi-agent AI system that implements RAG and MCP approach with different open-source LLMs. I want to talk about what kind of difficulties we face when we implement such a system. How to enforce security and data access roles? How to make the LLM generate correct SQL? How to deploy such a solution into an existing infrastructure?