Mark Needham
Clickhouse
Mark Needham is a Product Marketing Engineer at ClickHouse, where he works on short-form written and video content. He previous worked as a Developer Advocate at StarTree (Apache Pinot) and as a Developer Relations Engineer at Neo4j.Apart from writing blog posts and creating videos, Mark does developer experience, simplifying the getting started experience by making product tweaks and improvements to the documentation. Mark writes about his experiences working with all things data at markhneedham.com and has written two books on Graph Algorithms and Real Time Analytics. He also creates 5 minute videos on his YouTube channel @LearnDataWithMark.
We are both big fans of the live text commentary that the BBC provide for sports like football, tennis, rugby, cricket and more. While there is a lot of novel observations in the commentary, there is also a lot that is effectively summarising what just happened.
Wouldn't it be cool if the commentator could have a Co-Pilot who can make the process more efficient?
In this session, we will introduce an AI Co-Pilot for sports commentary based on Redpanda, ClickHouse, Flink, and a Large Language Model. A stream of events will be fed into RedPanda and we'll capture a window of those events on game-by-game and/or time-period buckets using Flink. These events, alongside historical match data, will also be stored in ClickHouse.
We'll then send the LLM the events that have just happened along with queries on historical data, from which it can come up with suggested text commentary. The commentator can then decide whether they want to use the Co-Pilot's suggestion, edit the suggestion, or just go along with their own version.
More Wouldn't it be cool if the commentator could have a Co-Pilot who can make the process more efficient?
In this session, we will introduce an AI Co-Pilot for sports commentary based on Redpanda, ClickHouse, Flink, and a Large Language Model. A stream of events will be fed into RedPanda and we'll capture a window of those events on game-by-game and/or time-period buckets using Flink. These events, alongside historical match data, will also be stored in ClickHouse.
We'll then send the LLM the events that have just happened along with queries on historical data, from which it can come up with suggested text commentary. The commentator can then decide whether they want to use the Co-Pilot's suggestion, edit the suggestion, or just go along with their own version.