OK, that’s the model – but where do you find the talent?
AI is not a new field but the demand is meaning skills are in short supply and there’s bidding war under way.
In the US, San Francisco – at the top of the Silicon Valley – is a city where employers are trawling most of AI-related skills. The shortage and the competition is pushing up salaries – an average of $157,335 according to Paysa.
Webroot’s Lonas says: “Much of the demand for these skills is coming from very high compensation companies and organisations, so it’s hard for small companies and startups to compete. My advice is to find one or two experienced experts, use them as the core of the team and then work with local educational institutions to find and fund programs.”
Think about using internships, special projects, and growing a farm team. “Think about hackathons and other non-traditional ways to find talent. Once you get critical mass, it’s easier because others will join knowing they can learn from your resident experts and add valuable experience to their resume and careers,” Lonas adds.
Sky Betting and Gaming has forged strong relationships with Leeds and Lancaster universities offering students work placements. Waterhouse says this is helping to remove some of the risk from the AI recruitment process. “It’s useful in getting people in. You can see what they’re good at and it gives them an opportunity to get up to speed with our business.”
Academia is a good place to start the hunt for ML experts – particularly those with a scientific and engineering background – but don’t rule out the self-taught. Contributions to ML-related open source projects or published research can be good indicators of technical ability. But prepare to invest in some upskilling, regardless of their background.
Wael Elrifai, senior director of Enterprise Solutions at Pentaho and the company’s AI and Machine Learning expert, is currently building a team of more than 20 engineers and data scientists. Having recognised that PhDs or Master’s degrees in machine learning are virtually non-existent, Pentaho has turned to training company Pivigo, which specialises in turning PhDs and MScs into Data Scientists and bridging the skills from traditional STEM degree areas to data science, machine learning and AI.
“Students have the opportunity during their training to apply what they learn by working on real projects. I recruited my last data scientist through a similar organisation and she is doing really valuable work for the team. She has a PhD in computational fluid dynamics, which has nothing to do with data science. After a four month conversion course, she now has strong practical knowledge in how to solve data science problems,” Elrifai says.
Bearing in mind how quickly the ML and AI fields are evolving, a proven ability - and a desire - to quickly learn new technologies is almost more important than pre-existing experience for members of your team. “PhDs are desirable but neither necessary nor sufficient; we've had great people without them and the occasional interviewee with them that made us wonder if they found it in a cereal packet,” Bickley says.
So technology is the key – right? Not quite and here’s where things get tricky. If it was a matter of simply finding qualified or aspiring data scientists and associated experts that make the task of building an AI team if not completely simple then at least relatively clear. Paysa found while 35 per cent of the open AI positions in the US required a Ph.D level qualification, 26 per cent needed just a master’s degree and 18 per cent a bachelor’s degree.
But what’s akin to gold dust in this hunt is finding people who possess a deep understanding of wider business. “It can be easy to get stuck in research-mode for a long time, and forget about the value of your work to the business. Your always need to make conscious decisions based on the data but also the cost/value analysis,” says Ocado software development manager Roland Plaszowski.
A killer combination is app developers who understand how AI/ML can give their product the edge, but who also have the ability to effectively collaborate with product managers who are closer to the customer.
“That will allow them to apply some intelligence to the usage data, learn about people’s habits and use that insight to develop a product that offers a smoother experience,” says inniAccounts’ Poyser.
So the team is assembled, but it’s not a thing that’s written in stone and the team’s composition will evolve.
During the early stages of your project, it’s likely data will dominate as ML engineers and data scientists will operate a full stack of analysis. Data scraping, cleaning and management can often consume a huge amount of effort.
As the project matures, so the team will grow and more specialized roles emerge, for example, with the addition of data engineers who manage big data infrastructure such as Spark.
You'll also find that the team follows a pattern familiar in traditional IT, particularly software development and DevOps.
"The differences between generic projects and ML/AI projects are not so big. We work hard to make sure that tests, continuous integration, monitoring, automation and documentation are in the project from the beginning, just like any other software engineering project,” Plaszowski said. ®