Data scientists, artificial intelligence experts and machine learning developers are in hot demand right now — so much, that these are some of the hardest jobs to fill.
According to this year’s best jobs report from Glassdoor, data scientist was the best job in the U.S. It was also the top job in 2017, and in 2016, up from ninth place in 2015. The number of job openings on the site rose from 3,449 in 2015 to 4,524 this year. And IBM predicts that the number of openings for U.S. data scientists and similar advanced analytical roles will reach 61,799 by 2020, with a 93 percent predicted growth rate in data science skills, followed by machine learning with 56 percent predicted growth.
And as demand increases, the number of candidates available to fill those roles is showing signs of not being able to catch up. According to Bloomberg, job postings for data scientists on Indeed.com increased by 75 percent between January 2015 and January 2017 — but job searches only went up by 65 percent.
It’s no wonder that enterprises are having a hard time finding qualified people. According to the 2018 State of the CIO report, 36 percent of respondents say that filling business intelligence and data analytics roles will be difficult, second only to cybersecurity. Artificial intelligence also made the top ten list, with 18 percent of respondents saying they anticipate AI role being hard to fill.
“There’s a significant demand for people with expertise in AI and machine learning,” says Tom Mitchell, professor of machine learning at Carnegie Mellon University.
Instead of hiring new employees, some companies are looking to help existing staffers get the advanced education and training they need to work in the areas of data science, artificial intelligence, and machine learning.
The most advanced training, such as the master’s program at Carnegie Mellon, takes between a year and a year and a half to complete, and requires knowledge of statistics, and basic programming skills. But there are also many online programs and courses that people can take, Mitchell says.
Here is a look at how several companies across a wide range of industries are training up current employees for the AI era ahead.
Back to school
Los Angeles County knows what the job market is like right now.
“We are also seeking resources that are hard to find when it comes to AI and business intelligence and automation experts,” says Murtaza Masood, assistant director at the LA County Department of Human Resources. He was previously the CIO of the department, and is leading many of the digital transformation initiatives for the county.
Recruitment efforts include partnerships with all regional academic institutions with relevant programs, and an internship program to create a pipeline of talent.
But the county has also been growing its own AI experts.
“Three years ago we established a center of excellence for AI that allowed us to start gathering existing resources within the county that had some expertise in this area,” he says. Employees interested in the subject can get tuition reimbursement for outside coursework, as well as in-house training.
Employees are particularly interested in algorithm design, he says, but the types of training employees pursue varies depending on the need.
“Business initiatives lead to platform decisions which lead to increased demand for training around those particular elements,” he says. Fortunately, the county has in-house talent to draw on. “Historically speaking, we’ve been early adopters of statistical analysis packages.”
For example, the county has recently been investing in monitoring and analysis tools to spot a variety of personnel issues, as well as potential cybersecurity problems. With 111,000 employees — LA County is the largest local government organization in the country — there’s a lot of data that can be analyzed to spot potential problems. But Masood wants to see machine learning and artificial intelligence used for more than spotting problems.
“Where I’m really excited, both as an HR professional and a technologist, is the ability to get training resources to employees to preempt all of that. The ability to serve up timely and bite-size information that enables an employee to improve their thought processes or behavior or knowledge, that is where I think the real power is and that is where the future might be,” he says.
Bringing training in-house
Capgemini is also in the middle of a big push right now to hire people with AI expertise. The consulting firm doesn’t write software, but it works with clients to configure and integrate existing software, and AI is now a major area of focus.
In addition to hiring experts, and sponsoring existing employees to get additional training, the company also has a digital acceleration center to help employees learn about new AI tech.
“We have a sandbox environment and we’ll trial the software in a laboratory,” says Tom Ivory, Capgemini’s head of strategic innovation. “We’re experimenting with dozens of AI technologies right now, and not every single one of them will be greenlit to move forward.”
At that point, Capgemini will train executives and delivery managers and other people on the team that works with customers on how to use those tools, he says.
The company also uses AI in its own operations. For example, Capgemini uses technology from automation vendor UiPath, and IBM Watson, to help process resumes of potential hires.
“We typically used keyword searches,” he says. “By using AI, there are more nuances that can increase the success rate of that individual being staffed on a project.”
At first, Capgemini used vendor-provided training, but soon there was a critical mass of internal experts.
“We had a baseline of management who were trained, as well as people who would actually have their hands on the keyboard, strategists, and folks who could understand how this technology would affect business processes,” he says. At that point, Capgemini brought the training in-house, using purchased training models from the vendor as the base.
Three years ago, the number of people at the company who trained in this technology was in the dozens — now, there are thousands, he says.
Embracing AI companywide
For tech companies that have an AI component to their products or services, employee training is even more critical.
Salesforce, for example, has been adding intelligence to its online customer relationship management platform. That means that in-house employees have to know their way around the technology. There’s an online training platform, part of the same Salesforce Trailhead system that’s available to the public.
“We have our own Trail, in which we have all this learning material that’s just freely available, that allows people to learn how to consume the AI capability that Salesforce provides,” says Marco Casalaina, vice president of product management of Salesforce Einstein at Salesforce.com. “It cuts across the company — we want everyone to be able to add intelligence, and to use intelligence, in applications in Salesforce.”
For example, an IT development team working on a room scheduling application for facilities management added intelligence to their application, to assign rooms to those customers who would benefit from the meetings the most.
“They’re not data scientists,” Casalaina says. “And our training is made for people like that, people who don’t have a data science background but want to make predictions.”
Cybersecurity firm Stealthbits Technologies is also investing in AI and machine learning skills. Some of that is from outside sources, says Jonathan Sander, the company’s CTO. “But the bulk of it is peer-to-peer.”
The actual training is different for different engineering groups, he says. “For the R&D folks who will get hands on, they are looking for the more comprehensive starter courses followed by the peer-to-peer materials we generate, and finally move into a mentoring system we have for building R&D talent. In that track they cover everything from the basics all the way through to application.”
On the support, consulting, and pre-sales side, the employees need to be able to understand how to use the technology and communicate its value.
“Our customers always benefit from more skilled engineers,” he says, “And the ML training has prepared them to give them good advice in using our ML-powered solutions as well as generally approaching ML as a valuable tool.”
Many companies think that machine learning is too complex or obscure to learn — that’s a mistake, he says.
“Ignore those impressions and get your hands dirty,” he says.
Like Stealthbits, big data firm Insight Engines also see a lot of value in peer-to-peer training.
“We start by incrementally exposing new staff to small areas of our development pipeline, where the changes they make have large-scale impact on our customers,” says Darien Kindlund, vice president of technology at Insight Engines. Then they move on to working on improving workflows, while still being supervised by experienced mentors.
So far about 65 percent of employees — both technical and non-technical — have received this kind of training, with a focus on machine learning, natural language processing, and data science, Kindlund says.
“By providing on-the-job training focused on customer impact in an applied setting, we have found our team absorbs, learns, and builds upon these skills much faster than in an academic setting,” he says.
Cybersecurity vendor Vectra Networks is also focusing on on-the-job, peer-to-peer training.
“Oftentimes, we have experienced mentors paired with new hires on projects to offer suggestions and guide them with machine learning approaches,” says Kevin Ni, the company’s data science team lead. “By training employees here the way we do, we can guide their development focused towards the unique problem space we are in. This means that customers can get results sooner. We also encourage employees to audit major online classes if they are unfamiliar with specific subjects.”
Expecting employees to learn on their own isn’t fast enough, says Joseph Kucic, CSO at Cavirin Systems, a cybersecurity vendor. Until recently, he was with the enterprise AI working group at Verizon.
“On the job, partner-mentor speeds up the ability to transfer knowledge and execute on the business goals,” he says.
If the need for AI talent is extremely urgent, then you can take radical actions, like acquiring talent by buying another company, he says. “But for longer term, the focus needs to be on a partner-mentor approach.”
Training at the pace of change
When training up employees in any emerging technology, keeping up with the rate of innovation is key. That means taking an ongoing approach to fostering new skills.
Invoca, a California-based call intelligence vendor, released its machine learning product, Signal AI, about a year ago. Since machine learning is now a core focus of the company, training is critical.
“We hold educational sessions across departments to educate internal teams about AI foundations, as well as specifically related to our AI solution offering,” says Sean Storlie, the company’s director of product management. There are quarterly AI-related trainings from the company’s data science team. In addition, employees can expense online tutorials, conferences, and onsite training.
“You can’t over-educate on this subject as things are changing so quickly in this industry,” he says.
“We create software that is supposed to also help other companies implement ML and AI,” says Lalith Subramanian, vice president of engineering for security, discovery and analytics at OpenText. “Internally, what we find is that the skills constantly have to be kept up to date.”
The company has a tuition reimbursement program for employees and also has been running training programs for the past three or four years, he says. “For instance, if we have a topic like commutative neural networks to be trained on, we bring in a training company and have groups of employees get the training,” he says.
But will they leave?
Some companies are reluctant to invest in AI-related training because they fear that the employees will immediately walk out the door, translating their new skills into big paydays elsewhere.
“In the tech industry, especially in Silicon Valley, a certain group of opportunists will get the latest training to look good in interviews,” says OpenText’s Subramanian. “But it’s not any more than if we didn’t have the training.”
In fact, in his experience, having the training available helps with retention, he says.
Another cybersecurity company offering extensive AI-related training to employees is Demisto.
“We are not worried about retention,” says company co-founder Rishi Bhargava. “As a company, we firmly believe that better trained employees are happier employees and do their jobs better. Moreover, encouraging upskilling is just one facet of our overall company culture. If a company helps an employee with training and the employee then leaves, the employee must not have been very satisfied with the job in the first place.”
Ongoing training also helps with retention at Aetna, the country’s third-largest health insurance company.
“It teaches them skills that are growing in marketability, and that’s a good thing for a professional,” says Jim Routh, the company’s CSO.
Aetna has a program to teach data science fundamentals to every one of the company’s security professionals, he says. By sticking with Aetna, employees not only continue to get this training, but they can also use the skills they learn on leading-edge applications, which means state-of-the art, unconventional security controls, he adds.
“We partner with a data science team of several hundred data scientists on our health side and our business side,” Routh says. “Together, we’re developing a core curriculum using a combination of commercially-available online training, and some training that we have developed in-house.”
So far, 200 security professionals have gone through the mandatory foundations class. “We’ve been teaching the course for two years now,” he says. “At first it was optional, but it quickly became the most popular course we teach.”