Oliver is a talented data scientist that often comes to work with bright ideas.
He likes to code, solve puzzles, and produce models, scripts and pipelines.
He spends many weeks of developing, training and building his code behind his laptop. This is what Oliver does best. At a certain moment, after a lot of testing, he knew that his code was ready for production. He was proud of what he had built. A true data-driven application.
The next day his manager came into the office with a big smile. “This is great!“, he said. “Can you bring it online, so we can start making use of it?”
That day Oliver started looking around how to bring his developed code into production. After Googling for hours on production, Python, Pytorch, pipelines, “how to deploy?“, Docker, VMs, it started to dazzle him.
“What about tools?“, he thought. Tools that could help him to serve his Python code with the ability to scale up without running out of compute? Tools that are easy to use? Tools that could log and monitor his code running in production? After days of searching and testing, Oliver realized there was not such an easy tool available.
This was not what he signed up for and not his area of expertise. He wanted to work on a new project and get back to data science.
Oliver was no quitter. It took him several months to build a serving infrastructure, installing libraries, managing dependencies, deploying his scripts and models, and creating an endpoint as a scalable microservice. After bringing his code online, within a month the infrastructure broke down.
This story was us.
That’s why we built UbiOps. So you can focus on data science, while we tackle productionizing your code.