Artificial intelligence jobs - The usual suspects
Artificial intelligence is a relatively new field. At least in a business context. As a result new job roles are appearing and it seems like there’s some confusion to the different roles and what they do. In an effort to shed some light on this confusion, I wanted to list the arch types, or the usual suspects, I meet in AI relations and give my view on their responsibilities.
Before you get started. Notice that the job roles here are strictly put. Of course reality differs and some of these roles are hybrids taken on by the same people a company often depending on company size.
Data scientists
The job title is somewhat self explanatory. It’s the science(or magic power) of taking in raw data in often huge quantities and identifying the signals in all the noise. The data scientists will often approach a project with trying to figure out what story the data is trying to tell and if there’s even a story at all. Finding out if the necessary data is sufficient to solve the problem for the project or if more or different data should be collected. So the data scientists are trying to understand and prepare the data for machine learning. Often there’s a lot of value in this process alone since a lot of aha's will appear in this process prevailing something about the business or the problem that wasn’t known before but was hidden in the data.
Machine learning engineer
The machine engineer is tasked with taking the data to an actual model that can work in production. Their goal is usually to have a high accuracy or precision as possible given some physical limitations. The limitation might be training time/cost or classification time/ costs. It could also be limited data. So the keyword for machine learning engineers is really efficiency. They get the most from what they have.
The approach is usually making some hypothesis about what standard machine learning algorithms would be the best (Maybe in a combination) and try it out. Once something starts to work, there will be a lot of effort to tweak and tune the algorithm to perform the best for the given problem at hand.
Product manager
The product managers in AI are tasks with the same tasks as usual product managers but it comes with some twists and special requirements. As usual the goal for the product manager is to make sure the product achieves its business goals and becomes profitable.
AI stands out on certain parts compared to traditional IT. If interested I wrote another post about the unique challenges of AI. In short the challenges are that you can’t predict the outcome(business value) or the cost of making the AI before you get started. So your business plan and your cost benefit will be more guessing than calculating. Data is also an equally important part of the product and that gives the product manager an extra dimension to work on. Lastly the user experience also often occurs from deep in the AI and should be managed with care. A good example is recall and precision that will mean a lot to the end users experience.
Data operations manager
Alright. I’ll be honest here. It’s not a role I have seen very often. I have seen the need in a lot of companies though and in Paperflow, we have a similar role.
The responsibility of the data operations manager is to make sure data is consistently being collected to train and keep the AI up to date with the world. It should be done by aiming for the highest quality at the lowest possible price. In many ways a efficiency problem like that of the machine learning engineer.
As written in an earlier post the data is often the biggest competitive advantage and not the ai itself. So don’t overlook this important position. Data is the fuel of any AI company and overlooking this could be a costly affair.
Data labeller
I wanted to include the data labeller as well. The people responsible for actually labeling the data. In many AI businesses I have visited this is the biggest department people wise. The task here is often to look through a lot of data and label it so the AI has something to learn from. The role here is important since this is where the quality of the data and as a result the AI stems from. Developing strategies and infrastructures in the IT to make this job as efficient as possible is extremely important for a business perspective. It is often overlooked but when doing AI projects this should actually be the first step to investigate. When you know how data can be collected and at what price then you get the first rough idea of what you are going into.
