The three AI roles in companies

I am currently halfway through the fantastic (and challenging) course put together by Jeremy Howard and Rachel Thomas with a tagline of “making AI uncool again”, I couldn’t resist digging in.

The the deeper I get into it, the more I’m realizing that perhaps there is an imbalance of emphasis of who we need to be building products incorporating AI. It’s not just the data scientists and PhDs, but we need the Engineers and Product people who can build out the whole pipeline.

When I was in High School, I was having to decide between taking an Engineering or a Computer Science degree. As a 17 year old with little exposure to the industry, it was hard to know the difference. When talking to professors in each, there was one professor who made it really clear to me the difference (which in my experience has been helpful):

“Computer Scientists build algorithms and architectures. Engineers leverages what the Computer Scientists build to make products and solutions”

With that, I became an Engineer and it’s been a great fit!

The AI commercialization roles

With the widely touted successes in Machine Learning in the past few years, the number of companies focusing on AI is exploding. Currently, the most sought after individuals in this space seem to be PhDs in the Machine Learning space. While these individuals are great at what they do, they are probably NOT the right fit for the majority of roles.

Data/Research Scientist

This individual is up to date with the latest ML architectures, they organize the data, prototype and build the models in an offline environment they are comfortable with and come up with the core model (algorithm) that worked best with the data provided.

Do Data Scientists need PhDs to be effective in this role? It seems that the answer is no. Looking at the top 25 ranked Kaggle competitors (data extracted here with some data unavailable), based on their LinkedIn they have completed PhDs (3), Masters (10) or Bachelors (6) and even no post-secondary at all (1). Wow!

In fact, the recommended requirement for starting this course is having 1 year of coding experience and some high school math!

Machine Learning Engineers

There needs to be to people that take the prototype model and can develop a framework to make it production ready. With the resource intensive learning components that Deep Learning possess, they need to be creative on how to manage the data pipeline, optimize the model for the cloud, manage the data for long term strategic purposes and manage the continuous learning that the model will need to do.

Product Managers/Designers with AI expertise

We still don’t know what our products will look like in the next 5 years, but if the current game changing products (self driving cars, Alexa, etc) are indicators, AI will be a common thread. How then will Product Managers be designing these products? This is a whole other topic but since having competitive data will be the foundation for the AI business model, a data strategy will be fundamental that the PM needs to think about:

  1. How can we quickly get some relevant data to prototype an AI product?
  2. How do we collect the training data at scale for the AI engine?
  3. How we store data for new products we don’t even know about for later use?
  4. How do we manage all the issues (marketing, positioning, 3rd party sales, anonymization and privacy) that arise?

In Conclusion

As you work through how you want AI to impact your product, think deeply about the holes that exist in your implementation strategy. It turns out they might not need a PhD at all!