Skip to content

Details

This month, returning presenter and long-time supporter of the forum Mark Rittman, CEO at Rittman Analytics will tell us about "Teaching AI to Work Like a World-Class Data Team"

Coding agents such as Claude Code can now build data pipelines, write SQL transformations and create dashboards in minutes. But speed is not the same as getting it right. Left to themselves, they guess at requirements, skip tests and produce work that is hard to review or maintain.
At Rittman Analytics we hit these problems early. Our answer was Wire, a framework that gives AI agents the structure we give a new consultant: defined project stages, agreed standards for data models and code, automated checks at each step and a person who approves the output. It runs inside the AI tools we already use and draws on a shared registry of our delivery processes. It has since grown into hosted agents that join our project teams, each with its own access to GitHub, Jira, Slack and the client’s cloud data platform.
This talk walks through how Wire works, using examples from real projects on BigQuery, dbt and Looker. We cover what improved, what went wrong and which decisions we still keep with people."
For reference, here's the docs and the Claude Code plugin and a few blog posts that describe what I'll be talking about.

We will be warmed up by our own Ashley Oldershaw, who is talking about IoT sensor data.

Have you ever built something assuming that a low-frequency annoying chore will barely ever happen, only for it to immediately happen twice? That's what happened as soon as I finished our migration from InfluxDB to BigQuery for our IoT data at Naked Energy, and we found out a sensor had been misconfigured and we had to correct its historical data.

We are a small team covering a lot of ground, so data pipelines that work only get touched when we have a really good reason to. This means that making a recoverable tool is absolutely crucial so it doesn't distract us while we are working on other stuff.

We went from a transform-on-read setup that coupled our data storage to our Django instance, to a transform-on-write setup that meant we could use BigQuery independently. We immediately got buyer's remorse when we had to correct a bunch of historical readings twice, and I was staring the solution in the face while teaching medallion architecture and dbt at Silicon Brighton's EVOLVE [26].

After this talk you'll have an idea of how to build a data pipeline with BigQuery and dbt with minimal setup, excellent recoverability and the knowledge you won't be making the same mistakes as me.

If you are stretched thin or starting a big migration, or just like hearing stories, this talk is for you.

Please be punctual, the event starts with networking and refreshments at 17:45, with the talks and live stream starting at 18:20.

The event is hosted by our loyal hosts at Barclay's Eagle Labs in their Platf9rm offices.

The Brighton data forum is supported by Silicon Brighton and sponsored by Rittman Analytics.

Related topics

Events in Brighton, GB
Business Analytics
Business Intelligence
Data Analytics
Data Management
Self Service BI

Sponsors

Rittman Analytics

Rittman Analytics

Rittman Analytics is a local data consultancy

You may also like