Almost every technology company openly talks about the shortage of personnel, but as soon as you send your resume to a seemingly suitable junior position, you encounter a paradox: everyone needs a specialist with minimal work experience, but experience can only be gained at work.
Here arises the question: what should novice specialists do?
For 9+ years, as the CEO and Founder of Data Science UA, I’ve been closely observing how the market and candidate expectations evolve. I see both sides at once: hiring AI specialists for her own team while helping other companies find theirs.
Today, I’m ready to share with you some tips that will lead to the most desired first offer.
.png?locale=en)
Start with the basics instead of looking for a “magic button”
The first year in AI is full of challenges and confusion, so to be sure of your future, you should move from building a solid foundation to practice, and therefore, start with mathematics.
It isn't enough to simply know what basic models do. You need to be able to explain why they work that way, what assumptions they make, where they break down, and how this affects the choice of architecture, metrics, and optimization methods.
For example, when you know that linear regression assumes feature independence, you can immediately see where this assumption breaks down when working with real data and why your metrics are falling.
At the same time, learn programming language, visualization libraries, and data/model handling skills. From the very beginning, create small but complete pet projects: the same emotion classifier in tweets or a movie recommender, which will consolidate your knowledge and also demonstrate the skills you have acquired. It allows you to experience the full cycle from data to production model.
Understand what skills are required
A team from the other side needs to see if you think like an engineer, not a student, or if you know how to understand new approaches, find the causes of problems, and turn insights into real improvements. Here's what they test for in a technical interview:
1. Analyze new approaches
In reality, you should understand fast what assumptions the model makes, what data it needs, how much the calculations cost, and where errors may occur. To pump up this skill, make it a habit: choose one new research paper per week, read only the abstract and results, and then try to explain the point in your own words.
2. Don't “patch symptoms” — fix the causes
Instead of “fixing” accuracy drops, you build data slices, look for edge cases, log metrics, and add profiling to clearly see where the model breaks. Based on data, investigate edge cases, expand the pipeline with logs/metrics/profiling, and ensure reproducibility to clearly identify the source of the problem. Take any dataset from Kaggle, intentionally “spoil” it (e.g., add noise or shift labels), then analyze why the performance dropped and document your process.
3. Ability to turn insights into improvements
It's easy to see that the data is noisy. However, coming up with a way to clean it up without losing useful features is a skill. After the changes, compare the results by metrics (eg, F1 or ROC-AUC). This is how you show real influence.
Find courses for professional growth
1. Machine Learning (Andrew Ng, Coursera)
Balances understanding of mathematics, intuition, and methodology. After completing it, you will not merely “know the formulas” but understand why they work. Also, it discusses the Numpy, Pandas, and Scikit-learn libraries, and provides reminders of the most important elements of linear algebra and mathematical analysis.
2. Deep Learning Specialization (DeepLearning.AI)
Trendy practices: about how to use modern generative models such as ChatGPT and how they work in general terms.
3. Full Stack Deep Learning / fast.ai
It focuses on real-world tasks in areas such as computer vision and natural language processing, and uses the fastai library to simplify the process of creating and training models with minimal code. Deploy, monitoring, updates, and how not to break down the production.
However, no course can think for you. Daily mini-projects, analyzing your own mistakes, and curiosity develop skills faster than any certificate (although a certificate upon completion of courses on LinkedIn is no less important).
If you understand what goes on inside your future job, what tasks, data, and results are involved, you will stop learning at random.
Search for cases to solve atypical situations
No matter how many tutorials you complete, real projects will surprise you. However, only these projects will help you develop the flexibility of thinking before you face atypical situations.
You may encounter corrupted data, unclear requirements, or infrastructure that doesn’t align with your model.
My team’s most unusual task was the visual quality control system. Algorithms confuse technical noise (black spots, reflections, background heterogeneity, visual camera artifacts) with real defects in the client's products.
The solution was found in the preliminary processing of images, specifically in the normalization of lighting and the elimination of system artifacts from the camera. This was exactly the case when even a small change provided a tangible improvement for both business and users.
Let the world know that you exist!
The previous point logically leads to the development of networking. Initially, it is important not to hide. Even the best knowledge won’t work if no one knows about you. Sometimes it is a coincidence, an acquaintance, a comment, or a post on social media that becomes the opportunity that opens the door to your first company.
- First and foremost, write honestly about who you are and what you are looking for. Don't be afraid to use different platforms for promotion: if you decide to use LinkedIn, indicate “open to work” in your status, talk about the courses you have taken, and the field you are interested in. On Facebook, join groups related to your specialization and respond to posts about internships or volunteer projects. On DOU, subscribe to the Junior and AI communities. Here, you’ll find the opportunities for newcomers.
- Once you have a basic understanding of the market, it's time to move on to the next step – writing a letter directly to the company. Ask about internships, projects, or offline schools. At first, you work to gain experience, and later, it’s your experience that does the work for you.
- Attend all themed events, lectures, and conferences, even if you don't understand anything yet. There are always people there who can give you advice, guidance, or introduce you to someone you need to know.
The future of the AI engineer
As for me, the future of an AI engineer will look completely different in just a few years: there will be less ‘AI for the sake of AI’, and more for real results.
The AI engineer will gradually become the system-level AI engineer: someone who builds functional solutions from ready-made models, tools, and services.
The main difference won’t lie in what kind of AI you use, but in the quality of your data and how quickly you can integrate a model into a product that truly delivers value.
At the same time, the role of agentic approaches and on-device solutions will only grow. AI will operate locally, respond faster, and rely less and less on the cloud.
Understand why you should be chosen for the position
There are at least twice as many candidates for each AI vacancy. Therefore, any employer needs to understand why you should be chosen among all the other candidates.
Although let's be honest: even if you're already on this path, don't expect to be hired right away. You'll still come out ahead at every interview:
- You will learn about new tools, resources, and books.
- You will hear how experienced people think.
- You will make an impression, and perhaps they will write to you again when you have grown to a level that currently seems “not yet yours”.
If you are offered a project, even on modest terms, accept it. This is your ticket into the profession. Only in practice can you see how your knowledge behaves in a real product, with live deadlines and mistakes.
Sooner or later, everything will work out. Remember that with every “no”, you are getting closer to the “yes” you dream of.
