How can I use machine learning for my business?
Have you trained a machine learning model?
Moreover, the question arrives, so now what?
Develop a front-end app with frameworks like Flask or Streamlit and deploy it as a service on the cloud.
Why should you deploy? because in real life, you don’t have CSV files to generate predictions in Jupyter Notebook.
You must think of data science as a glorified software development problem if you are already not thinking that way.
If you can’t productionalize your machine learning pipelines – they are of no use. Your 99% AUC is just a number nobody cares about.
If you are working for companies with well-defined teams, infrastructure, and data-driven culture, then chances are you will get a lot of support from engineers to productionalize models from your notebook.
However, in most cases as of today, there is a lot of ambiguity. Furthermore, companies are still looking for their first use-case, they don’t know what to hire. Who to hire? And if they are just starting out, chances are that they start with a very small team (as small as 1 person). In addition, if you are that person, you are expected to deliver end-to-end projects.
Learn implementation – You are competing with seasoned software developers, how hard do you think is for them to learn .fit() and .predict()?
If all data science problems were solved by running random scikit-learn ensembles using .fit(). Then we would indeed face a lot of competition from “seasoned software developers”. Luckily, analytics is much more than that.
