Dreaming of building a successful career in the AI ​​industry? Experts from Grammarly, MacPaw, SQUAD and more tell you what to look for and where to study

The path to AI today is markedly different from what it was 10-15 years ago. The pace of change and the number of tools create a completely new environment to start.

Now it is important for a junior specialist to combine fundamental knowledge, practical skills, and the ability to quickly adapt to new tools and models.

Does it still make sense to start in AI now?

Absolutely. AI remains a field without clear educational «paths.» However, the fact is clear: the demand for engineers who can combine technical expertise with a practical understanding of business challenges is growing faster than new LLM models are emerging. The global AI market is projected to reach $800 billion by 2030, reflecting its rapid adoption across industries and everyday life.

For those just considering a career in AI, now is the best time. The industry is in a state of active formation, and each new specialist can influence how the technology will work in real products and systems.

Where to start your journey in AI

Despite all the variety of starting points, experienced engineers are sure of the main thing: theoretical knowledge without practical application remains just beautiful concepts on paper. Professional skills are formed in the process of working on real tasks, where you have to deal with unstructured data, limited time frames and specific business requirements.

Oleksandr Romanko, Associate Director at SS&C Technologies Canada and a lecturer at the University of Toronto, is convinced that the most important aspect is awareness of the choice. «You should start by deciding whether AI is your calling. If you are sure that it is, then with the ability to solve practical problems using AI algorithms and make decisions based on modeling results. Understanding the mathematics of AI, algorithms, validation of calculation results (the code for calculations is generated by AI in most cases), validation and implementation of modeling results will also be useful,» he says.

Volodymyr Kubitsky, Director of AI at MacPaw, looks at the start from a different angle: you can’t do without a foundation. «I would definitely recommend starting with basic fundamentals — at least linear algebra and basic mathematical analysis to understand derivatives and matrices. This is just the groundwork that is needed to move forward,» he emphasizes.

Kubitsky recommends participating in Kaggle competitions in parallel with studying courses on Coursera: «It immediately gives a feeling of something more real than synthetic. And it gives an understanding of the non-ideality of the conditions in which real work will take place.»

Mariana Romanyshyn, Area Tech Lead and Computational Linguist at Grammarly, adds that learning should be done through small but concrete projects: «You choose a problem that can be solved with the help of AI. This can be your own idea or an existing solution that you are trying to replicate or improve — it is important that this task excites and motivates you. While you are mastering the field of AI, you build a basic primitive solution for the chosen problem, research similar/existing solutions, collect data, build metrics, and iteratively improve your own implementation, applying the acquired knowledge and skills.»

This approach will provide not only knowledge, but also personal experience with the difficulties that had to be overcome. This is what employers then value. Romanyshyn places a special emphasis on community: asking questions, looking for mentors, and being part of a professional environment is no less important than working alone.

As a result, a holistic picture emerges: some start with mentors and mini-projects, some with mathematics and competitive platforms, and some with internal choice. All three approaches combine into a practical formula: fundamental knowledge + lots of practice + conscious choice of profession.

Courses and trainings that will help an AI engineer grow rapidly

Oleksandra Boguslavska, CEO & Founder at Data Science UA, gives specific recommendations for those who like a clear plan:

1. To create a basic foundation in machine learning — Andrew Ng: he will help combine mathematics, intuition, and methodology to create a holistic picture of knowledge.
2. DeepLearning.AI specialization in deep learning to gain modern practices and experience the trade-offs between quality and speed.
3. A practical engineering course that covers the entire scope from model development to production deployment (like Full Stack Deep Learning or fast.ai) to understand how AI models work in real life.

Denis Lazarenko, Applied Science Manager at SQUAD, also notes Andrew Ng’s courses and adds another interesting platform: Stanford CS231n: Deep Learning for Computer Vision. «This is a must-have for those who want to become strong in computer vision. The course helps to understand the architecture of neural networks and their application in practice,» he says.

Volodymyr Kubitskyi advises not to limit yourself to specific names, but to build your educational path using the tools provided by the technologies themselves. «Everyone has their own path, and here Generative AI can become a navigator. Plan your path through programs, courses, and trainings based on AI agents that will help you choose to learn the skills that are needed in the labor market now and will be needed in the future.»

Three key skills that distinguish an AI engineer in the market

Oleksandr Romanko emphasizes the combination of technical and business competencies. «Understanding what you are modeling with AI, understanding AI algorithms, the ability to validate modeling results and implement them in practice. In addition to technical skills, business soft skills such as: problem solving, creative thinking, management skills, MLOps.»

Oleksandra Boguslavska shared what her IT recruitment agency pays attention to when selecting candidates:

1. Speed ​​of analysis of new approaches. Be able to decompose a new approach into assumptions, data requirements, computational cost, and possible types of errors in 1–2 hours, formulate several hypotheses, and confirm or refute them with minimal experiments.
2. A systemic approach to problems. Instead of locally eliminating symptoms, systematically build data slices, investigate edge cases, expand the pipeline with logs, metrics, and profiling, and ensure reproducibility to clearly identify the source of the problem.
3. Turning insights into improvements. It’s not enough to find a problem, you need to be able to transform the understanding into specific steps to improve the system.

According to Kubitsky, AI leaders are characterized by the following features:

1. The ability to quickly understand something new. In AI, everything changes every week, and when you are really into the subject, you don’t wait for someone to come and understand and put everything on the shelves: you take it, read it, try it, run it, test it with your hands — you form your own opinion and experience.
2. Make it work. Not perfect, not beautiful, but simply — work, because the data is always skewed, if there is any at all, the task is not fully understood, time is short and here it is important not to get stuck, but to achieve a result, even if it takes several iterations and is not ideal, but a result
3. Balance between accuracy and practicality . In AI, you always have to choose something: either more accurate, or faster, or less, and here it is important not to chase the best metrics just like that, but to understand: «What exactly do these users need in this product?»: sometimes it is speed at the cost of quality, sometimes it is the opposite.

The AI ​​engineer of the future: who is he?

AI is evolving so rapidly that five-year predictions seem like fantasy, but the main trends are already visible. According to a McKinsey study, the use of AI in organizations has increased from 55% in 2023 to 78% in 2024. At the same time, most organizations are using AI in several business functions, such as IT, marketing, sales and product development.

According to Oleksandra Boguslavska, in 5 years, an «AI engineer» will evolve into an AI system engineer. This means a shift from «architecture selection» to designing holistic solutions from ready-made models, tools, and services. At the same time, the industry is facing standardization of assessment, stricter regulation, energy and cost restrictions. As a result, a constant search for compromises between different requirements.

Mariana Romanyshyn emphasizes: «Making predictions is a very thankless task, especially for 5 years ahead. Therefore, I would rather say what I would like to see in the field of NLP in the next 5 years:

• large language models for national languages, adapted to the cultural and historical context;
• significant progress in detecting disinformation;
• integration of AI into educational processes: an AI tutor, adapted to the language, subject and level of knowledge of the child,
• generation of tests for self-testing, development of critical thinking and self-education skills through AI, etc. AI engineers definitely still have a lot of work to do.»

Denis Lazarenko notes that we can already see the growth of the capabilities of large language models (LLM), which is changing the very role of the AI ​​engineer. The ability to use AI tools effectively will be increasingly important: to delegate routine tasks to models, and to direct your time to system integration, work with the product and create value for users. In essence, every AI engineer becomes a little bit of a product manager.

At the same time, the technical base is not disappearing anywhere. Knowledge of programming and algorithms is needed to critically evaluate the results of models, detect «hallucinations» or incorrect code.

Another direction is the strengthening of the role of MLOps. While it becomes easier to create a model, it becomes more difficult to build a reliable, secure, and scalable infrastructure. And this is where the demand for engineers will only grow.