Data Platform Engineering
Thoughts and learnings from building scalable and reliable data platforms applying platform engineering principles. Disclosure, I’m a Google Cloud Champion Innovator
- Indexed issues, last 90 days
- 3
- Latest publication
- Aug 29, 2026
- Audience
- Checking…
- Earliest in this view
- Jul 25, 2026
Latest issues
How do you know if the analytics agent is right? (opens the original)
Read excerpt
So you’ve deployed an AI agent that answers analytics questions. People are using it, it returns numbers, everyone seems happy.But do the numbers actually check out?This is a real problem that I think many teams initially just skip over because the demo looked good and the stakeholders were excited. An analyst querying your warehouse brings years of context with them. They know which table to use for revenue vs bookings, they know to filter out internal test accounts, they know that “active cust
The aesthetic credibility trap (opens the original)
Read excerpt
There’s a failure mode in AI-generated analytics that I don’t think gets enough attention. The agent produces a beautiful, well-formatted dashboard: clean charts, clear narrative, consistent colors, a summary table with bold totals. And a third of the KPIs are wrong.The person reading it has no idea. The polish is indistinguishable from correctness.Visual quality creates an impression of factual quality, even when the two are completely uncorrelated. It’s arguably more dangerous than an agent th
From passive code reviews to active PR fixes (opens the original)
Read excerpt
Google’s decision to sunset the consumer tier of Gemini Code Assist on GitHub was the catalyst for this setup. While the initial notice was inconvenient, shifting away from standard automated comments turned out to be a solid upgrade.Rather than looking for another tool that simply posts review comments, I moved the workflow over to Google Jules, which is already included in my Google AI plan.The main difference is actionability. The code assist PR bot left inline suggestions that still require
Publishing over time
Last 90 days. Choose a month to open its work.
Recurring subjects
Named in the text we hold. One piece can cover several.
Audience
No verified audience measurement yet.
About this data
Counts cover the work we have indexed. Tone needs enough text and a confident classification. Excerpts and episode notes are not full articles or transcripts.
Identity or attribution wrong? Suggest a correction.