A Brief History of Machine Learning Platforms

Ian Hellström | 23 September 2020 | 3 min read

Data and machine learning technologies for the big data era have been developed in the last twenty years. Let’s review these two decades in terms of data processing and machine learning frameworks as well as machine learning platforms that have been spotted in the wild.

While ML platforms such as IBM’s SPSS and SAS have been around for decades, open-source data and machine learning technologies are a recent phenomenon. Modern machine learning (ML) and deep learning frameworks (for Python) have only been around for 10–15 years.

The associated end-to-end machine learning platforms have only popped up at tech companies in the last five years or so. No wonder frameworks and platforms have not yet converged towards a dominant design, especially since we have been able to witness advances in public clouds, containers and orchestration, DevOps in years since 2000. In other words, the entire cloud-native suite has not yet been around that long.

Bold entries in the timeline below are end-to-end machine learning platforms announced by the respective companies. As you can see, the first tech company to come forward with details of their platform was Facebook in 2016, a mere four years ago. Since then, many more have published details of their platforms.

If you prefer a shorter version focused exclusively on ML, please have a look at my 30-minute talk AI Chihuahua: Why Machine Learning is Dogged by Failure and Delays at the virtual Cloud Native Summit in October 2020, which is also available on YouTube.