Forum Navigation

What competencies are needed for machine learning jobs? How much of mathematics is necessary?


To create a great AI team, you need to understand the work content:

A project is defined by :

Technical objective: Machine learning is about determining or predicting an output by learning the pattern / features of the business problem. It is about regression (predict an analog value - ex: the temperature in server room on a given day and time, predict rental value) or classification (predict which product ad to show for best chance of buying among the set of possible options etc.).

ML project will have

1. A model:

2. The the way to feed values (input features) into it and take output (labels):

3. How the model is deployed and used.

4. Of course the data.: Tools to extract/create data and do pre-processing

Out of all, at least one ML expert. The following would be the least expectations :

a. Has delivered some software in any domain

b. Basic understanding of most of AI technques

c. Good programming skills

The second would be a good programmer. To be comfortable in the team, better know ML. Gain as much knowledge (for team acceptance) as possible. The main work is the data input pipeline into the model. One would also need to do data pre-processing.

The third person would be domain person who also has AI familiary. But the person needs to have lot more familiarity with the domain and should be able to say what is right and what is not wrt data. Lot of data would be required for training the model. If the data is not proper, then predictions also wont be satisfactory. The same person could do training and predictions and could gain familiarity with tools and techniques (for example: tensorboard which is used to understand how well the model is learning)

Later for deployment , there might be need for sophisticated cloud hosting for serving the model. If this is the need, then cloud architect, software architect or a combination of two might be needed.

The above is the most basic team. Depending upon the scale, the team could be bigger.

Few points to be clear:

You do not need a PhD in machine learning project team. You need software engineers and domain experts who have picked up knowledge of math/ ML libraries like Keras , Pytorch, Tensorflow etc. and also the domain. Experience of having delivered software in the past would greatly help.

Hence the following would be my recommednations:

1. A basic ML course which gives overall picture with relatively less pain. Andrew NG's course is famous.

2. Specializations - Overall ML and/or Deep Learning (Deep learning solves most problems, but is data and CPU/GPU hungry). Go for either Edureka Machine Learning or Andrew NG's specialization would do.

3. Programming language skill I suggest python.

4. Software architecture or cloud (For the relevant member).

5. Domain expertise

If you keep these things in mind, you can confidently face machine learning interview. If someone asks deep academic mathematics question, you can clarify what is really needed. If they are not convinced, there are better employers waiting for you.








Sharing is caring