According to the latest Global AI Jobs Barometer report, the number of U.S. job postings requiring artificial intelligence skills increased by 588% between 2012 and 2024. At the same time, the industry is undergoing a rapid transformation. Employers are imposing increasingly demanding and multidisciplinary requirements, while the proliferation of AI-powered coding tools is leaving few opportunities for entry-level specialists without hands-on experience. ML and AI engineering expert Dave Savostyanov explains where the market is heading, how it is expected to evolve in the near future, and how an innovation-driven scientific approach to problem-solving can help businesses address their most pressing challenges.
Dave Savostyanov describes the current market for ML and AI professionals as exceptionally dynamic. Demand for qualified talent is at an all-time high, fueled in large part by the rapid pace of technological advancement.
“Strictly speaking, there are also plenty of specialists actively seeking employment,” he explains. “The problem is that these are mostly entry-level professionals, or so-called ‘juniors.’ They are trying, with varying degrees of success, to master different tools, but they are not yet capable of coming in and solving a specific business problem, building a system, or improving existing models and the infrastructure around them.”
This is further complicated by the emergence of specialized AI-powered coding tools, whose capabilities are sufficient to take over many of the tasks performed by these junior specialists.
“To sum up, demand is growing faster than people can retrain,” Dave Savostyanov explains. “According to industry analysts, 2.5% of all job postings in the United States now mention AI skills, representing a 55% increase over the past year and nearly 300% growth over the past decade. In addition, the technology stack is evolving too quickly. The skills that were sufficient to be a competent and in-demand ML engineer in 2021 are no longer enough today. Many employers are no longer looking for specialists focused solely on machine learning. They need versatile professionals who can train models, collect data, build and maintain infrastructure, tackle frontier challenges, and more.”
Savostyanov points to his own field of expertise, namely the development of neural network models for motion generation, as a particularly illustrative example. One of the defining characteristics of this area is its broad range of applications, spanning everything from the film and video game industries to the design of humanoid robots with fine motor skills capable of interacting effectively with their physical environment.
“These technologies are in extremely high demand today,” says Dave Savostyanov. “When it comes to robotics, I firmly believe that within the next decade, robots will become as commonplace in our daily lives as microwave ovens or toasters are today. As for the entertainment industry, such models can give content creators entirely new capabilities by significantly accelerating the animation process and quite literally expanding the boundaries of what is possible.”
According to Dave Savostyanov, these kinds of challenges fall squarely into the category of business pain points that cannot be solved with a quick fix. Moreover, some of the practical cases he has worked on were essentially uncharted territory from the outset, with no existing references or ready-made solutions available.
“We have made a scientific approach a core part of our work,” Savostyanov says. “If ready-made solutions do not exist, they have to be developed. That means engaging with the scientific community: attending research conferences, following academic journals and publications relevant to our work, reading papers, speaking with their authors, and much more. In most cases, after carrying out an extensive review, we gain a fairly clear understanding of what we need and how it can work.”
The second stage is engineering work. It involves collecting the necessary data, deploying and configuring computing resources, and interpreting and analyzing the initial results. Meanwhile, metrics are developed to track improvements in model performance through clear and measurable indicators.
This approach, Savostyanov explains, is most commonly found in companies and startups operating at the cutting edge of science and focused on commercializing breakthrough research. Moreover, making such an approach the foundation of professional practice allows companies to respond effectively to business needs and deliver precisely the results required in each specific case.
As for the future of the labor market, the expert believes the outlook is mixed. On the one hand, Savostyanov says, AI systems have become so proficient at writing code that they are already capable of freeing up as much as 50% of programmers’ working time, allowing them to focus on more creative tasks. This shift in the development process has, in turn, reduced the cost of building digital products, making projects that were once considered too expensive to develop and maintain economically viable.
“You can now operate across different markets and countries and build vertical agents for virtually any purpose,” Savostyanov says. “As for specialists, however, they will have to adapt significantly. In the past, programmers were primarily responsible for writing code, whereas today they are increasingly expected to take on the role of software architects. They need to view systems holistically, connect different components, and assume greater responsibility. Additionally, despite the rapid development of AI-powered coding tools, the market itself will continue to expand, as new niches emerge and opportunities that were previously considered economically unfeasible become viable. Yes, there will undoubtedly be periods of disruption and setbacks, but those are natural parts of technological progress.”
