From 3a2c1b5976215c51e2b845543171485783ef6425 Mon Sep 17 00:00:00 2001 From: Donald Wade Date: Wed, 6 Nov 2024 12:39:17 +0800 Subject: [PATCH 1/2] Blog post on Joe Reis at DataEngBytes --- _posts/2024-11-01-dataEngBytes-2024-1.md | 69 ++++++++++++++++++++++++ 1 file changed, 69 insertions(+) create mode 100644 _posts/2024-11-01-dataEngBytes-2024-1.md diff --git a/_posts/2024-11-01-dataEngBytes-2024-1.md b/_posts/2024-11-01-dataEngBytes-2024-1.md new file mode 100644 index 00000000..480163b4 --- /dev/null +++ b/_posts/2024-11-01-dataEngBytes-2024-1.md @@ -0,0 +1,69 @@ +--- +layout: postv2 +title: DataEngBytes Perth 2024 +font: serif +description: A run down of this year's conference talks +date: 2024-11-07 +tags: + - community + - conference + - perth + - data + - data-modelling +author: Donald Wade and Sophie Giraudo +--- + +# Byting the dust: siloed approaches to data modelling + +> Miss out on this year's DataEngBytes conference? FOMO dialled to the max? Fear not. Mechanical Rock's intrepid team of conference-goers went along in person, so you didn't have to. + +Since its beginnings in 2019, DataEngBytes has established itself as a real highlight of the Australian tech conference calendar, and true to form, it brought some great speakers to Perth this September. + +_Donald Wade_ and _Sophie Giraudo_ digest Joe Reis's talk on the future of data modelling. +## Joe Reis: Mixed Modal Arts + +As well as being a world-renowned expert and teacher in all things data related, Joe Reis is a keen fan of Mixed Martial Arts, and his talk at this year's DataEngBytes used the evolution of MMA as a key metaphor to help drive a pretty compelling thesis. + +Back in the early '90s, when MMA first became a "thing", competitors could just about get by in the ring if they were masters in their own discipline - be that boxing, Brazilian Jui Jitsu, Muay Thai. + +But time moves on, and today, "just" being an expert at a single discipline just doesn't cut it. Not only do fighters need to know a _minimum_ of boxing, Brazilian Jiu Jitsu and Muay Thai - they need to know them _really, really well_. + +Even more than that, to be truly effective, they need to have internalised the techniques to instinctively know when to apply which technique when. + +### Wax on, wax off + +As it is in the world of Mixed Martial Arts, so it is in the world of data. + +Back in the day, it might have been enough for a practitioner to be a Kimball dimensional modelling or Data Vault expert, Joe argues. A software engineer might also be an expert at relational data modelling. And an ML engineer or data scientist would know the ins and outs feature engineering and deep learning. + +Historically, experts in different fields had a tendency to stick to narrowly to their own lane, unconcerned with how people downstream might need or want to consume their output. But things are changing. + +### AI is everywhere + +AI is everywhere. It will move outside and beyond the current hype, and it is guaranteed to become increasingly pervasive. Multimodal data is also becoming more and more common. And of course the corporate world now increasingly looking to use LLMs to interrogate their data. + +But, to quote Joe directly: "AI without good quality data means you'll do dumb things more quickly. And the underlying data model - combined with semantic structure (think knowledge graphs, semantic layers, etc) - is the key to making this work." + +>AI without good quality data means you'll do dumb things more quickly. + +### Data-powered products are everywhere + +A single software product today might rely heavily on OLAP processing to provide a better user experience for end users, and it might also use a Machine Learning model to drive another part of the customer experience. Depending on what you're using it for, the application data needs to be modelled differently. + +As a result, anybody who works with data today needs to understand how to model data across a number of different situations. + +The upshot is that roles within the world of data are evolving. Increasingly, practitioners need to know a wide variety of disciplines - and they need to know them really, really well - so that they know which technique to apply in which situation. + +The future is as exciting as it is multi-disciplinary. + +--- + +Joe is CEO of Ternary Data. He describes himself as a "recovering data scientist" and is co-author of the best-selling book **_Fundamentals of Data Engineering: Plan and Build Robust Data Systems_** (O'Reilly) and he co-hosts the popular podcast **_The Monday Morning Data Chat_**. + +As well as Joe's DataEngBytes talk, we've used his original blog post as reference for this article. Please read it [here](https://joereis.substack.com/p/mixed-model-arts-the-convergence). + +--- + +Mechanical Rock has extensive experience helping clients increase their confidence in data-driven decision making. Find our more or get in touch with us [here](https://www.mechanicalrock.io/our-expertise/data-and-ai). + +--- From 39bde2cdf8c65a00a37abc46ff737eccbbdfb9fe Mon Sep 17 00:00:00 2001 From: Donald Wade Date: Thu, 7 Nov 2024 13:53:11 +0800 Subject: [PATCH 2/2] Fix some typos. --- _posts/2024-11-01-dataEngBytes-2024-1.md | 26 ++++++++++++------------ 1 file changed, 13 insertions(+), 13 deletions(-) diff --git a/_posts/2024-11-01-dataEngBytes-2024-1.md b/_posts/2024-11-01-dataEngBytes-2024-1.md index 480163b4..481e4f86 100644 --- a/_posts/2024-11-01-dataEngBytes-2024-1.md +++ b/_posts/2024-11-01-dataEngBytes-2024-1.md @@ -13,22 +13,23 @@ tags: author: Donald Wade and Sophie Giraudo --- -# Byting the dust: siloed approaches to data modelling +# Byting the dust: siloed approaches to data modelling > Miss out on this year's DataEngBytes conference? FOMO dialled to the max? Fear not. Mechanical Rock's intrepid team of conference-goers went along in person, so you didn't have to. Since its beginnings in 2019, DataEngBytes has established itself as a real highlight of the Australian tech conference calendar, and true to form, it brought some great speakers to Perth this September. _Donald Wade_ and _Sophie Giraudo_ digest Joe Reis's talk on the future of data modelling. + ## Joe Reis: Mixed Modal Arts -As well as being a world-renowned expert and teacher in all things data related, Joe Reis is a keen fan of Mixed Martial Arts, and his talk at this year's DataEngBytes used the evolution of MMA as a key metaphor to help drive a pretty compelling thesis. +As well as being a world-renowned expert and teacher in all things data related, Joe Reis is a keen fan of Mixed Martial Arts, and his talk at this year's DataEngBytes used the evolution of MMA as a metaphor to help drive a pretty compelling thesis. -Back in the early '90s, when MMA first became a "thing", competitors could just about get by in the ring if they were masters in their own discipline - be that boxing, Brazilian Jui Jitsu, Muay Thai. +Back in the early '90s, when MMA first became a "thing", competitors could just about get by in the ring if they were masters in their own discipline - be that boxing, Brazilian Jui Jitsu, or Muay Thai, to name a few. -But time moves on, and today, "just" being an expert at a single discipline just doesn't cut it. Not only do fighters need to know a _minimum_ of boxing, Brazilian Jiu Jitsu and Muay Thai - they need to know them _really, really well_. +But time moves on, and today, "just" being an expert at a single discipline won't cut it. Not only do fighters need to know a _minimum_ of boxing, Brazilian Jiu Jitsu and Muay Thai - they probably need to know more martial arts - and they need to know them _really, really well_. -Even more than that, to be truly effective, they need to have internalised the techniques to instinctively know when to apply which technique when. +Even more than that, to be truly effective, competitors need to have internalised the techniques so they instinctively know when to apply which one. ### Wax on, wax off @@ -36,31 +37,30 @@ As it is in the world of Mixed Martial Arts, so it is in the world of data. Back in the day, it might have been enough for a practitioner to be a Kimball dimensional modelling or Data Vault expert, Joe argues. A software engineer might also be an expert at relational data modelling. And an ML engineer or data scientist would know the ins and outs feature engineering and deep learning. -Historically, experts in different fields had a tendency to stick to narrowly to their own lane, unconcerned with how people downstream might need or want to consume their output. But things are changing. +Historically, experts in different fields had a tendency to stick to narrowly to their own lane, unconcerned with how people downstream might need or want to consume their output. That wasn't great. But things are changing. ### AI is everywhere -AI is everywhere. It will move outside and beyond the current hype, and it is guaranteed to become increasingly pervasive. Multimodal data is also becoming more and more common. And of course the corporate world now increasingly looking to use LLMs to interrogate their data. +AI - or at least talk of AI - is now ubiquitous. It will move outside and beyond the current hype, and it is guaranteed to become increasingly pervasive. Multimodal data is also becoming more and more common. And of course many in the corporate world now want to use LLMs to interrogate their data. But, to quote Joe directly: "AI without good quality data means you'll do dumb things more quickly. And the underlying data model - combined with semantic structure (think knowledge graphs, semantic layers, etc) - is the key to making this work." ->AI without good quality data means you'll do dumb things more quickly. +>"AI without good quality data means you'll do dumb things more quickly." ### Data-powered products are everywhere A single software product today might rely heavily on OLAP processing to provide a better user experience for end users, and it might also use a Machine Learning model to drive another part of the customer experience. Depending on what you're using it for, the application data needs to be modelled differently. -As a result, anybody who works with data today needs to understand how to model data across a number of different situations. +As a result, anybody working with data today needs to understand how to model data across a potentially wide variety of different situations. -The upshot is that roles within the world of data are evolving. Increasingly, practitioners need to know a wide variety of disciplines - and they need to know them really, really well - so that they know which technique to apply in which situation. +The upshot is that roles in the world of data are evolving, and they will continue to do so. One thing seems certain, though. Increasingly, practitioners need to know a wide variety of disciplines - and they, too, need to know them _really, really well_ - so they don't inadvertently bring a knife to a gunfight. -The future is as exciting as it is multi-disciplinary. --- -Joe is CEO of Ternary Data. He describes himself as a "recovering data scientist" and is co-author of the best-selling book **_Fundamentals of Data Engineering: Plan and Build Robust Data Systems_** (O'Reilly) and he co-hosts the popular podcast **_The Monday Morning Data Chat_**. +Joe is CEO of Ternary Data. He describes himself as a "recovering data scientist" and is co-author of the best-selling book **_Fundamentals of Data Engineering: Plan and Build Robust Data Systems_** (O'Reilly). He also co-hosts the popular podcast **_The Monday Morning Data Chat_**. -As well as Joe's DataEngBytes talk, we've used his original blog post as reference for this article. Please read it [here](https://joereis.substack.com/p/mixed-model-arts-the-convergence). +As well as Joe's DataEngBytes talk, we've used his original blog post as reference for this article. You can find it [here](https://joereis.substack.com/p/mixed-model-arts-the-convergence). ---