A Hallmark of a ML Practitioner
Difference between Theory & Practice
As we move along in the Knowledge Economy, knowledge, skills and experience will become more important as they provide the fuel for the company to move forward with new ideas and innovations. It is the reason why I constantly ponder on the questions, “How can we measure skills, proficiency, and competencies?” Or “What makes someone a practitioner and not a paper theorist?”
This is a problem, companies especially their HR department will have to tackle if they want a trustable performance grading system. In case, you are wondering, nope I have not come to something ground breaking but if I do, you will be the privilege few to know. :)
However, I do think there is one hallmark that I feel separate “experts” from the practitioners. We need it because the industry is flooded by “experts” whom if you look at their background and experience closely, has never code, train models or even test the ML model or applications. The most absurd I have seen is, “AI Ethics expert”, has not implemented a ML model at all. Of course, its absurdity is just my own opinion.
Coming back, I am a strong proponent of implementation/application because that is where all the value of a data projects are realised!
Before any implementation of ML models into the business process, there is a lot of testing to be done. This to me is the hallmark of a practitioners versus the paper theorist, who has never done any implementation and has such never participated in the testing at all.
Practitioners will know how to carry out testing, through experience, very quickly dive into where possible coding bugs are, where data might have gone wrong, check the hyperparameter testing and more importantly, write out test cases that should “break” the model calculation and/or business processes.
Just to give an example: “Being able to write a test cases for a credit scorecard model, where the final score sits on the cut-off point and see how does the credit approval system handles it.”
Through how someone write test cases and test the ML model calculation and its subsequent process, we can truly see how good a practitioner someone is. :)
Your thoughts? :)
To continue supporting my work, consider sharing this and get me a “book” or two over here. :)
Thanks for reading till the end! I hope this has sparked more thoughts in you!