The dearth of AI and data talent is a conversation starter that is often mentioned at conferences and roundtable discussions. And this shortage is only going to get more acute, as demand soars while schools and training courses struggle to churn out enough employees trained in AI and data.
Taking on coursework
A recent article in Fortune offered some insights into how Amazon is tackling this talent crunch – by requiring potential new programming hires to take classes in machine learning. This was shared by Bratin Saha, a vice president and general manager of machine learning at Amazon.
According to Saha, company executives at Amazon think they can teach these developers ML basics over “a few weeks” so that they can work on more cutting-edge projects after being hired.
This entails teaching developers Python, as well as rudimentary ML concepts that include statistical regression and deep learning. The former are techniques used for making highly accurate predictions over time, while the latter revolve around training neural networks.
It is worth noting that recruits are not graded as they would in a formal college course. Instead, the courses are designed to give developers a foundation in statistics and ML to better equip them with the ability to understand the theory behind them. The test, so to speak, is to create either a recommendation system or forecasting model.
With a good grasp of how ML works, these new developers are hence much better equipped to create new AI products or troubleshoot AI-centric systems. It is not clear if those who do poorly will still be hired, though taking this course comes late in the hiring process – presumably only after underachievers are weeded out.
It is also interesting that Saha is seeing an increase of university students in the US who are enrolling in AI courses. This mirrors what is happening in the Asia Pacific region, as universities and employers are increasingly offering ML and data science courses.
“To train or to hire” is a common dilemma that we hear executives fret over. At Amazon at least, the solution appears to revolve around doing both.
Image credit: iStockphoto/oatawa