Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Lars Albertsson 🌻🇺🇦 Profile picture
Lars Albertsson 🌻🇺🇦
@lalleal
I help companies extract value from data - https://t.co/teanMCmfJz. Data factory engineer, 1st edition data bard. @lalle@mastodon.technology. Tech-only tweets: @mapflatcom.
Jul 17, 2021 • 6 tweets • 1 min read
This, so much. Most important data tweet ever. This is the most important factor predicting ability to use data and thereby AI.

Few companies are organised around value. Many clients have come to me and wanted data and AI technology. 1/6 I have talked to them, worked with them, but cannot twist their organisation. They get tech, but little value. 2/6
Jan 25, 2021 • 41 tweets • 13 min read
@gpadres Not scalability, but many reasons. I should write a blog post "SQL considered harmful for data pipelines", since the tradeoffs are poorly understood. I am not good with blogs, but I guess this is a good excuse to write an MVP of that post in the form of a twitter thread. 1/100 @gpadres A proper blog post would have example code to make a better comparison, but I lack available time right now. Perhaps I'll make a sequel (sorry for pun) one day.
May 10, 2020 • 6 tweets • 2 min read
I find it fascinating how the cloud providers' different approaches to this issue reflect their organisation and culture.

The problem with the combination of weak object store consistency and ETL pipelines is known since Hadoop started appearing in AWS, i.e. ~10 years.

1/n Amazon is organised in autonomous "two pizza" teams. Holistic aspects tend to fall between cracks in autonomous cultures, and cross-cutting concerns are difficult to address. In autonomous cultures, incentives tend to reward launching new things.

2/n
Mar 6, 2020 • 9 tweets • 9 min read
@_J_sinclair @HoloMarkeD @GCPcloud I have helped many companies on this journey. Some unsolicited advice:

1. Strive to get from a push-based workflow (fill the lake) to a pull-based. I.e select use cases of business value and ingest the data they need into the lake. @_J_sinclair @HoloMarkeD @GCPcloud 2. Take use cases all the way and show value before embarking on new use cases.

3. Implement only what the use cases need, but first paint a clear long-term goal picture. Each step should take you towards this goal.