Disrupting machine learning
87% of the projects never even go in to production.
You are picking ripe Swedish strawberries but could only eat them in 6–8 months?
That’s how machine learning operations are done at companies today. To go from an idea to data collection and have that go in to production is a process that takes time. A lot of problems are in the discrepancy between data scientists, analysts, data engineers and DevOps. Imagine you are building a house where the carpenter and the architect are blindfolded and speak different languages.
Creating a Jupyter notebook environment, adding Git, installing some libraries into a container and calling it a solution that you charge money for , should be a crime.
— Petter Hultin Gustafsson, founder of MLOps
All the hype
Let’s face it, everyone wants to involve the data they have with AI but 87% never even go in to production. It’s mostly cool keynotes that give insights with old data. Connecting this gap has been why the concept MLOps has become more relevant today. Companies wants to solve todays problems with todays data and not be stuck in this mess.
MLOps will be the benchmark in how to apply machine learning in companies soon and even now platforms like MLOps, Valohai and DataBricks are showing companies how to plug in their data and do what used to take months in minutes.
If the time it takes machine learning to produce impact is significantly lower I believe it can help us solve many problems faster. If machine learning can become cheaper and faster it will become more accessible for smaller actors and organisations to apply it on their data, the control of it will no longer be in the hands of a few.
MLOps is available as a plug and play solution to your AWS account -