Azure Machine Learning feels like Windows Phone
…and I mean that as a high compliment.
I know I’m not alone. I loved my Windows Phone. My wife and my kids all loved theirs too and we don’t care who knows it! …but I digress. My focus here is on developer tooling – not my consumer experience.
Mining, curating and cleaning data that’s already sitting out there somewhere waiting to be found is the new gold rush.
If you have a compelling idea for an AI/Machine Learning app – don’t be intimidated. Sure, there are powerful tools available from Google, IBM and others – but Microsoft has, once again, put together a development platform and tooling that takes out (nearly) all of the startup friction you might expect. Here's what worked for me - maybe it'll be useful for you (or a friend).
Nearly a decade ago, I had been immersed in application analytics best practices and tooling – and I needed a way to get into the weeds with a real project. The only tiny little issue was that I hadn’t done any coding in years. Microsoft was in full-swing evangelizing mode and promoting Windows Phone development and so I dove in. After installing Visual Studio, I watched a few videos, started with a few samples, used Telerik UI controls and, in a couple of weekends, I had launched the first of what ultimately turned into a family of mobile yogi apps with over ¼ million users (my wife is the yogi and David Poeschl seriously improved the engineering in subsequent apps) – but the point is that I was able to do it and along the way I learned quite a lot about analytics patterns, privacy issues, licensing models, and more (here’s a post from “back in the day” Marketplaces Matter and I’ve got the analytics to prove it).
…but, to put a finer point on this, if I had had to rely on the tooling and vendor support that came with iOS or Android, my first “hello world” would never have seen the light of day.
Fast forward to 2019, and I’m now “immersed” in Machine Learning best practices and tooling – and I need a way to get into the weeds with a real project. …and Microsoft is in full-swing evangelizing mode and promoting AI/ML development – so I took the dive.
I created my Azure account.
I didn’t need to install anything locally this time – I’m using Azure Machine Learning Studio.
I watched a few hours of MITOPENCOURSEWARE starting with Introduction to Machine Learning
Bought and sped through UDEMY course A-Z Machine Learning with Azure Machine Learning
…and in a couple short days, I’ve launched my first real – WORKING – Web Service (what I mean by this is that my trained model can solve a novel, potentially valuable, problem. I'm not talking about completing a sample - it's wholly new. …and along the way, I’m learning quite a lot about ML dev patterns, processes and workflow that I think I would have most likely missed if all I did was read about development versus doing.
…and much like my experience with Windows Phone, if I didn’t have easy access to Microsoft’s platforms and tooling, I most likely would never have even launched a “hello world.” THANKS MSFT!