Finding a job destination in Data Science

In switching from Science to Data Science you want a sense of destination: Where will you be working?

Switching careers is a bit more elaborate than switching jobs, and you need more information than that you will be swapping your ≅€40k p.a. postdoc position for ≥€60k business role, with the realistic expectation of reaching ≅€100k p.a. in a senior role (in less than five years). In this post, I will take you through the following considerations:

1. The ‘Da Vinci’ moment in Machine Learning enabling fundamental domain switching.
2. A systematic search for the greatest opportunity by identifying domains with high demand and low entry barriers.
3. Developing a matrix listing your preferred domains as well as the approaches or tools most interesting to you.

The ‘Da Vinci’ moment. Much of machine learning is fresh, and most approaches are generic rather than specialized. Indeed, machine learning currently is a transferable skill ‘par excellence’. Some think that this is principally so and point to, for example, DeepMind as a single AI, or to the search for the master algorithm (Pedro Domingos). On the other hand, some are working on making machine learning easier, faster, and more reproducible (‘machine teaching’). In this scenario domain expertise becomes increasingly important, and specialization ensues.

What I see currently is that people entering Data Science can switch domains freely and successfully, e.g. from life science to automotive, chemistry to energy, engineering to music, and so on. Thus, when switching from science to data science you have the chance to re-evaluate your interests and re-calibrate your trajectory. If you feel like making a change, possibly even switching career track as well as domain: You can do it.

The greatest opportunity. Finding the first (new) job is never easy. Anyone hiring typically prefers candidates with experience. An also not uncommon scenario is that the hiring manager may want you but, for example, human resources may oppose an offer. On the other hand, I have noticed that PhDs joining Data Science training programs get offers before they have completed the program.

How to identify high demand and low barriers? If a topic is prevalent in the media (e.g. autonomous driving), and a provider emerges promising to find and train new talents by the thousands (e.g. Udacity’s self-driving car engineer program), this is a fair indication of high and accelerating demand. Udacity’s program amounts to about 500 hours of training, indicating also that entry barriers currently are low. Other data to consider include startup funding trends, job advertisements, and industry trends indicating that AI is becoming a core feature of the product.

Your search matrix. What I am trying to tell you with the above reflections is that you can develop your search matrix uninhibited. You don’t have to look at your current skills set and then narrow your search to the domain to which you think they are transferable most easily. The ‘Da Vinci’ moment, and the increasing number of fields with high demand, mean that you enjoy much greater degree of freedom than you may think initially.

So here is what you can do: Browse industries and domains for the their ‘ML readiness’: Which product ideas and customer problems interest you? Look across approaches in Machine Learning and experiment with the tools: Which are the most promising? Where are you the most competent?

You will obtain a matrix, on which you can pinpoint which combinations of domain and approach are most attractive to you. That is where I would start the journey to Data Science and Machine Learning.