Machine learning and AI have been a dominant skill for years and organizations understand is a requirement to be competitive, especially in 2021 to help scale their business. Machine learning can help in various capacities including deep analytics, pattern recognition, natural language processing, automation and much more. Now take these powerful skill sets and combine them with programming and yet another powerful skill, cloud computing, and you have somewhat of a unicorn for business scalability
For decades we have relied on spreadsheets to serve as our analytics platforms while we bring in analytically minded people who could ingest and manipulate datasets into the likes of MS excel and maneuver the data so it can be put together in a tidy presentation that summarizes hours or even weeks of work. While we still continue to rely on spreadsheet, the intent has shifted and more organizations are starting to see the value behind leveraging the machine to do the thinking and number crunching and people to still help interpret the results.
So why hasent everybody jumped on the ML bandwagon and stopped their legacy practices? Well it takes a lot of computational power to run modern day machine learning algorithms. While large organizations have access to CPU’s, large GPUs and even TPUs, small and medium sized organizations don’t have that horsepower just yet. For example when google did made their transformer model public, the training was done over months using 8–10 large GPUs along with several scientists to accomplish a production ready open source framework.
So how do organizations that don’t yet have a large infrstructure benefit from this? In for the win comes cloud computing that provides organizations access to enterprise grade hardware and architecture that can help them leverage and scale this equipment at the push of a button. For starters google has a service called google colab which enables you to run your very own Jupyter like notebook and leverage googles GPUs, CPUs and TPUs for free (with restrictions of course) or opt for a paid version for more features. The benefit is organizations can scale their needs up or down on an as need basis vs investing is costly architecture. It also gives them a platform to run their machine learning modules without the need to install any software.
Check out this YouTube video that introduces the concerto of machine learning using google colab:
Machine Learning in the Cloud for Free | Google Colab
Now if you’re in for something a bit more robust and permanent, you can always leverage more mature cloud platforms including google cloud platform (GCP), Microsoft Azure, or Amazon AWS. These platforms offer full solutions from infrastructure services to platforms and even full blow pre-trained models like text analytics
I would also add to the list some other alternatives including Heroku, Digital Ocean and Linode as servers that can be easier scaled up or down to suite your business needs
If your organization is looking to get into ML and learn more about how they can leverage machine learning with cloud computing be sure to check out out levers.ai