Table of Contents
- There’s a shortage of machine-learning talent, recruiters and industry specialists say.
- The shortfall stems from more businesses requiring artificial-intelligence expertise.
- Recruitment experts, academics, and those who changed careers explain how to pivot to the industry.
Machine-learning specialists are highly sought after right now.
Recruiters and experts told Insider they were facing an acute shortage of machine-learning skills as the demand for specialists in artificial intelligence moved beyond tech and into sectors such as healthcare and finance.
Machine learning is a commonly used form of artificial intelligence that involves the use of self-learning programs and algorithms. It underpins a lot of services, from the movies
recommends to fraud detection for banks. The technology allows computers to process and draw patterns from huge amounts of data, which makes it useful in a variety of fields.
In a national survey of businesses conducted in June by the UK’s Department for Digital, Culture, Media, and Sport, about one-quarter of respondents reported a shortage of machine-learning skills.
The hiring market is competitive for qualified candidates. Analysis of US disclosure data on foreign-labor hires in 2021 shows base salaries for machine-learning engineers ranged between $73,000 and $250,000, with a median of $152,125. European and UK salaries, however, tend to trend lower.
With the demand for machine-learning engineers outpacing the supply, Insider spoke with recruitment experts, academics, and machine-learning late bloomers to find out the top tips for those looking to pivot to machine learning.
1. You don’t necessarily need a Ph.D., but prepare to work hard
While most machine-learning engineers come from highly academic backgrounds, the number of roles now requiring machine learning skills has helped open up the jobs market.
“There will be a class of roles that require top-level skills, probably people who’ve done Ph.D.s and had that very academic route,” Matthew Forshaw, a senior advisor for skills at the Alan Turing Institute, said. “But the vast majority of the 238,000 roles that the UK needs are not those.
“There’s a middle ground where you don’t need to know the statistical foundations of absolutely everything to be able to identify which models are appropriate in which setting. It’s a bit sector-dependent and depends on the size of the organization.”
Universities also can attest to the change as companies scramble to recruit grads with machine-learning skills.
“Historically speaking, most people probably went on to do Ph.D.s, rather than directly going into industry,” Mark Herbster, the program director for UCL’s Master of Science in machine learning, said. “There’s some shift there. We have many more students going directly into the industry and startups.”
Ivan Lobov, a research engineer at DeepMind, studied public relations and advertising at Moscow State University, before working as a corporate strategist at a digital-marketing company. He had been interested in computers since childhood but didn’t pursue this passion until much later in life.
“I didn’t understand what questions to ask or where to find guidance,” Lobov told Insider.
He started taking vacations to participate in weeklong hackathons and competed in online competitions set by Kaggle, a data-science-community tool owned by Google where participants hone their skills through challenges.
“After years in the field, I think I’ve covered most of the gaps in my education to a level where I think it’s hard to tell I don’t have a STEM background,” he said. “But it was tough sometimes.”
2. Find ways to learn on the job, or in your spare time
For anyone hoping to emulate Lobov, he said it was important for wannabe machine-learning engineers to “find approachable tasks that motivate you.”
“I found Kaggle to be the most useful tool,” he told Insider. “But don’t aim to be a grandmaster. Use them to motivate you to learn more skills — to go into nitty-gritty details of the algorithms you’re using.”
Lobov’s colleague Deeni Fatiha, a product manager on DeepMind’s applied-AI team, previously worked in material sciences, researching everything from how plastics can be made more biodegradable to the use of fiberglass in construction projects.
“I had no formal background in machine learning or computer, so I had to learn a lot from scratch while on the job,” Fatiha said. “I would keep a running list of all the things that came up that I wanted to learn more about and would read up on them in my spare time.”
Franki Hackett recently won the “rising star in tech” award at the CogX Awards in London in recognition of her work at the AI firm Engine B, which is applying machine learning to accounting and auditing services. But she wasn’t always set to be a techie.
After earning a bachelor’s and a master’s in politics, Hackett worked in communications for a string of different organizations and nonprofits in London. She was then accepted to a graduate scheme at the UK’s National Audit Office, which employs data scientists and researchers who help assess the financial decisions and policies made by different government departments.
“The more I looked into it, the more fascinating I found it,” Hackett said, adding that she was able to learn on the job by “picking the brains of all the experts there when I needed to.”
She eventually became a lead data-analytics manager for the organization before getting recruited to become the head of audit and ethics at Engine B.
Hands-on experience is one of the best ways to gain technical skills. The Alan Turing Institute’s Forshaw recommended “embedded, incubator-style training and working with domain experts to get the in-the-room experience” when reskilling from other disciplines.
For those who go down the university route, this means placement opportunities and collaborating on projects.
3. Whatever your background, don’t be intimidated
Khyati Sundaram started her career in finance, working for JPMorgan and the Royal Bank of Scotland before pivoting to machine learning.
“I specialized in mergers and acquisitions, but after six years in the industry, I was itching to do something different,” she said.
After earning an MBA from London Business School, Sundaram launched a startup, Fosho, to help make supply chains more sustainable via AI. She learned the basics with the London School of Economics and Political Science’s online course Machine Learning: Practical Applications.
“Overcoming others’ doubts was by far the biggest challenge,” she said. “I knew I was capable of mastering machine learning and AI. But as a woman in business, particularly in tech, those around me had other ideas.”
Sundaram is now a cofounder and the CEO of Applied, an AI-powered hiring platform designed to help employers remove biases from their recruitment processes.
“Don’t be put off by the hype and all the talk around how hard it is to work in this field,” Engine B’s Hackett said, adding: “There is a lot of hype around artificial intelligence and machine learning.
“Yes, there are tricky concepts and challenges, but it’s not magic. It’s not beyond you. Find people who can explain things in simple terms — these are normally the best people to help you learn and grow.”
4. An unconventional background can work to your advantage
“Transitioning from a different sector can also be a great advantage. People with varied job histories come with a whole host of transferable skills,” Sundaram told Insider.
DeepMind’s Fatiha agreed, telling Insider there was a real need for a “diversity of perspectives” in the machine-learning space.
“We need people of different backgrounds in terms of demographics but also in terms of professions, to help inform the powerful solutions we build with ML,” she said.
“Some of the most insightful conversations I’ve had at DeepMind have been with research scientists who have backgrounds in medicine, the performing arts, and philosophy,” she added.
As for the reskilling process, candidates shouldn’t count themselves out over a lack of technical experience.
The Alan Turing Institute’s Forshaw said commercial and creative skills could be equally important.
“People who’ve got strong domain expertise or strong commercial acumen from a route other than computing or statistics or one of these conventional feeders into data science,” he said, adding: “I find it easier to reskill them on the technology than I would to try and fill in the other part for someone who is deeply technical.”
Forshaw said: “If we fast-forward two years, a lot of the technology will have changed. But the things that are permanent are professional values around ethical and trustworthy use of data, communication, and relationship building.”