Quiet(er) holiday season days are perfect for reflecting on the previous year, and making plans for the upcoming 2020. What kind of an impact are you going to make next year, and what are the tools and solutions available in the market today to help you get there as fast and efficient as possible?

When those goals involve the use of AI, I often hear the first objective is to build a data science team. I disagree. Read why!

I was inspired by a year old writing by George Seif titled "Don’t make this big machine learning mistake: research vs application”. Seif makes a distinction between machine learning research a.k.a. science and machine learning application a.k.a. engineering.

Science

If your focus is research, you will be hiring PhDs, who attend machine learning conferences like the one just hosted in Vancouver. They submit papers, write patents and are generally amazing in creating some novel approach that takes AI as a field one step further. If this is your path: you are probably a very large company, you have a big budget and AI innovation is your business.

Just look at the 1427 papers published in the Vancouver conference, called NeurIPS. This research will benefit us all in the coming years, and I admire everyone involved in it.

Engineering

If, instead, your objective is to apply machine learning to win new business, solve a problem within your operations, or generally just beat the competition; you will probably make a mistake if starting with a scientist hire. What you need is an engineer who can use the right tools for your problem and get results fast.

We were all moulded into thinking that scientists are needed for any type of machine learning applications. Just a few years ago there were barely any machine learning SaaS (or MLaaS) tools available in the market, and getting anything done required a scientist who was able to build and optimise models using open source libraries.

That’s not the case anymore. It is an amazing time to be an engineer!

Approximate global population of data scientists and software engineers

In my day to day work, engineers use Aito to predict missing fields in processing purchase invoices with RPA - implemented in just a few days. Or check how an engineer at Intwixt uses Aito to create intelligent workflows in Slack - again in just a day. Or how IKEA built an extremely complex demand simulator prototype in 5 weeks, again using Aito. Imagine the amount of time and $$ saved compared to reinventing the science!

We are not alone at Aito. Look at the machine learning related announcements AWS made at their re:Invent event in Las Vegas, or the options Google Cloud AutoML already offers. To see an example how to put these in use, check Bruno Amaro Almeida’s talk on "what can Serverless AI/ML APIs tell us about Fake and Credible News” where he walks through the usage AWS Comprehend for sentiment detection.

2020 is the year when engineers will use machine learning tools that are easy, fast to use and affordable. I have seen the struggle of hiring data scientists, and the time+cost it takes to get useful results, even after the team is in place for >6 months. An engineer with the right tools will help you reach your goals not only faster, but also in an agile way, giving you great visibility on the process. And, there are about 10 software engineers for every data scientist.

Now, who do you hire in 2020?