Hiring Data Engineers and Data Scientists in Belarus: Skill Map and 2026 Demand

A head of data at a US fintech spent ten weeks looking for a “data engineer / data scientist hybrid” in Belarus. After ten weeks of mostly bad fits, we walked her through the actual market: those are two different roles, the hybrid she described doesn’t really exist at scale anywhere, and the budget she had would land her either a strong DE or a strong DS — not both.

We see this pattern often enough that it deserves a proper guide.

This post explains: how the four data roles actually divide up in the local market, what each one’s stack looks like in 2026, current salary ranges, and the sourcing channels that actually deliver. It’s the same conversation we run on every data hire call.

The four data roles in 2026 (and why hiring managers still conflate them)

Most companies asking us for a “data scientist” actually need a data engineer. A meaningful share of the ones asking for a “data engineer” sometimes need an analytics engineer. And the AI engineer category that emerged in 2024–2025 has now eaten part of both DS and ML engineering. The first job of any data hire is figuring out which of the four you actually need.

Data Engineer (DE). Pipelines, warehousing, lakehouses, streaming infrastructure. Owns the data platform end to end. The most-asked-for role in 2026 and the one most often mis-titled by hiring managers.

Analytics Engineer. dbt-shaped middle ground. SQL-heavy, transforms data for analyst consumption, builds the semantic layer between raw ingestion and downstream consumption. A smaller specialty locally but growing fast. Often filled by DEs transitioning into the role rather than dedicated AE candidates.

Data Scientist (DS). Classical statistics, modeling, experimentation, A/B testing, causal inference. The role that has shrunk most in relative demand as the AI engineer category emerged, particularly at the senior level where the strongest candidates increasingly relabel themselves.

ML Engineer / AI Engineer. Productionizes models, owns MLOps, increasingly does LLM and RAG work. The fastest-growing and highest-paid of the four in 2026. The pool of candidates with real production LLM experience is genuinely small everywhere — Belarus included.

The practical implication: before you write the job description, decide which of these you need. “We need someone who can do all of it” is the request that produces ten weeks of bad fits. Pick one, write the role for that one, and budget for the right level.

The Belarusian talent pool: where it’s deep and where it thins

An honest mapping of the local market by role.

Data engineering is the deepest specialty. Wargaming, banking analytics, and the agency ecosystem (EPAM, IBA, Itransition) have been training DE talent for fifteen-plus years. Mid-to-senior data engineering is the strongest segment to recruit from. Junior DE is also available if you’re willing to train.

Analytics engineering is smaller but growing. Most candidates with strong dbt experience came through DE before transitioning. Dedicated AE candidates exist but you’ll see fewer of them, and they cluster around companies with mature data platforms.

Classical data science is solid at mid-level, thinner at the principal level. The strongest senior DS candidates increasingly position themselves as AI engineers because that’s where the money is. Pure statistics-and-experimentation roles see fewer applicants than they did three years ago.

ML / AI engineering is the segment where supply has tightened most. Strong production LLM and RAG experience in particular is scarce. Most candidates who have it are already employed, and a meaningful share are being actively recruited internationally. You’ll pay for senior LLM-experienced engineers, and you’ll wait longer to find them.

Senior English fluency is historically stronger in data engineering than in data science, partly because DE has been more international-client-facing through the agency ecosystem. ML/AI engineers vary widely. Worth screening explicitly rather than assuming.

The skill map: what’s actually on the resumes

Grouped by role. Names tools as category examples without endorsing them.

Data Engineer skill map

Languages. Python and SQL are universal. Scala still appears on older Spark codebases but is declining; if it’s on a CV, it usually means a multi-year project the candidate is trying to move on from.

Cloud. AWS is dominant (about 70% of senior DE CVs). GCP is growing fast, particularly at companies founded in the last five years. Azure shows up at enterprise-leaning shops.

Warehousing. Snowflake and BigQuery split most modern stacks. Redshift on legacy AWS-heavy projects. ClickHouse on analytical workloads.

Lakehouse and storage formats. Databricks (Delta Lake) and Apache Iceberg both present. The 2025–2026 shift is that Iceberg has eaten more ground than expected — table format experience now matters as a discrete skill on senior CVs.

Orchestration. Airflow is the default. Dagster and Prefect appear on newer projects. Look for actual production orchestration scars on the CV, not just “familiar with”.

Transformation. dbt is the default for SQL transformation. SQLMesh appears on a small minority of newer projects but isn’t yet a market expectation.

Streaming. Kafka is universal where streaming matters. Flink for serious low-latency work. Kinesis on AWS-heavy stacks.

Data Scientist skill map

Languages. Python is dominant (Pandas, NumPy, scikit-learn). R appears on academic-background CVs and is declining.

ML frameworks. PyTorch is dominant. TensorFlow appears on older projects, declining. JAX shows up on research-leaning CVs.

Statistics. statsmodels and scipy are standard. PyMC for Bayesian work, which is increasingly relevant for marketing-mix and causal-inference roles.

Experimentation. Strong DS CVs show clear A/B testing experience and a working understanding of causal inference. Vague “ran experiments” without a methodology is a signal to dig in during the interview.

Notebooks and experiment tracking. Jupyter is universal. Databricks notebooks are common in lakehouse shops. MLflow is the most common experiment tracker; Weights & Biases is widely used in deeper ML shops.

ML / AI Engineer skill map

Productionization. MLflow, BentoML, KServe, and custom Kubernetes setups. Production experience is the dividing line between mid and senior here; everyone has trained a model locally, fewer have served one to real traffic.

Feature stores. Feast appears most often. Tecton on more enterprise-leaning setups. Hopsworks occasionally.

LLM and RAG (the 2026 differentiator). LangChain (declining toward LangGraph), LlamaIndex, vector databases (Qdrant, Weaviate, pgvector), embeddings work, retrieval evaluation. Strong candidates have hands-on with at least one open-weights deployment (Llama, Mistral, Qwen) alongside the obvious closed-API providers.

MLOps tooling. Kubeflow on more mature setups. Airflow + MLflow combinations remain common. ZenML on newer projects.

Profiles on Hugging Face (public models, datasets, Spaces) and Kaggle (competition history with medals) are higher-signal for ML and DS candidates respectively than anything on LinkedIn — worth checking before the first interview.

2026 salary benchmarks

Two tables, in EUR gross monthly, fully loaded for a Belarus-resident hire. Adjust for currency, model (EOR vs. own entity), and additional benefits.

Data Engineer

Level	Years	Monthly range (EUR, gross)	Notes
Junior	1–2	€1,500–2,300	Mostly from agency programs or junior analytics roles
Mid	3–5	€2,500–4,000	Deepest part of the DE market
Senior	5–8	€4,000–6,500	Where most international hiring concentrates
Lead / Principal	8+	€6,500–9,000+	Platform owners, multi-team scope

Data Scientist and ML / AI Engineer

Level	Years	Monthly range (EUR, gross)	Notes
Junior DS	1–2	€1,400–2,200	Mostly PhD or bootcamp pipeline
Mid DS	3–5	€2,400–3,800	Solid market depth
Senior DS / ML Engineer	5–8	€4,000–6,800	ML Engineer at top of range
Lead / Principal / AI Engineer	8+	€6,800–9,500+	LLM / RAG specialists at the upper end

Practical notes on top of the numbers.

The AI engineer category has bid the senior range up through 2025–2026, particularly anyone with production LLM, agentic, or RAG experience. We’re seeing offers at the upper end of the Lead / Principal range for engineers with two solid years of LLM production work, even when their nominal title is still ML Engineer.

Junior DS roles are getting harder to defend at companies investing in AI engineers instead. If you have a junior data hire budget in 2026, you’ll often get more value from a junior DE who can grow toward analytics engineering than from a junior DS in a market where the senior DS role they’d grow into is itself shrinking.

The DE-to-AI-engineer transition pays better than the DS-to-AI-engineer transition because the infrastructure baseline is closer. Worth knowing if you’re advising current employees on career direction.

HTP residency keeps personal income tax at 9% versus the standard 13% — a meaningful effect on take-home pay, particularly at senior levels. The HTP guide covers the tax and operational implications in more depth.

Sourcing channels: where to actually find data talent

Every channel has a job. Here’s what each one is good for, and what it isn’t.

jobs boards. The standard for the local IT pool, including data roles. Maintain an employer page if you’re hiring more than one data role per quarter.

Habr Career. Particularly strong for data and ML. Historically the platform where Russian-speaking data engineers and scientists have been most active. Profiles here are often more current and more detailed than the same candidate’s LinkedIn.

ODS (Open Data Science) community. Russian-speaking community of data scientists and ML engineers. Sourcing through it directly is a local-recruiter skill, but it’s worth knowing about — much of the senior DS and ML conversation in the region happens here.

Kaggle profiles. Higher signal for data scientists specifically. A meaningful Kaggle history (medal-level competition results, recent activity) correlates with practical modeling skill. Lower signal for DE or AI engineering — a strong DE doesn’t need Kaggle, and a strong AI engineer is usually too busy shipping production systems to compete.

Hugging Face profiles. The closest equivalent for ML / AI engineers. Public models, datasets, and Spaces are high-signal for hands-on LLM work.

LinkedIn Recruiter. Works for senior DE and ML engineering. Less effective for mid-level — many strong data engineers maintain sparse LinkedIn profiles. The data scientist community is even sparser on LinkedIn locally.

GitHub. High-signal for DE and AI engineering. Filter by contributions to relevant projects — Airflow operators, dbt packages, ML or LLM tooling. Lower volume than job boards, but the signal-to-noise ratio is materially better at senior level.

Telegram channels. Active data-specific and ML-specific Russian-speaking channels are underrated and shift constantly. A local sourcer who tracks the current landscape adds real value here.

Specialized recruitment. When you’re hiring more than one data role in a quarter, agency partnership shifts the ROI math, especially for AI engineering where the search runs longer. Our IT recruitment service handles DE, DS, and ML / AI engineering as core specialties.

Rough channel mix for a typical senior data search: 30% job boards (dev.by, Habr Career), 30% direct sourcing (LinkedIn + GitHub + Kaggle or Hugging Face), 25% network referrals (the local data community is small and knows itself), 15% specialized agencies.

Job Description anti-patterns that lose you data hires

Five mistakes we see every week.

1. “Data scientist who can also build pipelines.” This is the unicorn from the cold open. Pick one role and write the JD for that role. Asking for both gets you neither at the level you actually need.
2. No mention of stack. “Modern data stack” is a phrase, not a JD. Name the warehouse, the orchestrator, and the key tools. Strong candidates filter on this before deciding to apply.
3. Requiring Kaggle Grandmaster status for a senior DS role. Filters out strong practitioners with no competition history. Kaggle is one signal, not a credential.
4. “5+ years of LLM experience.” LLMs were not in production five years ago. JDs that ask for impossible histories filter for nobody. If you need senior LLM experience, the realistic ask is 1–2 years of production work.
5. No SQL test in the interview process. Every CV claims SQL fluency. Many don’t deliver at the level a senior DE or analytics engineer needs. A 30-minute live SQL exercise catches this fast and is the single highest-ROI screening change we recommend.

Realistic hiring timelines

From recent placement data, broadly:

• Mid Data Engineer: 4–6 weeks from kickoff to signed offer.
• Senior Data Engineer: 6–8 weeks.
• Mid Data Scientist: 5–7 weeks.
• Senior Data Scientist: 7–10 weeks.
• Senior ML / AI Engineer with production LLM experience: 10–14+ weeks. The supply is genuinely thin.

These timelines assume the JD is right and you’ve decided on the operating model — EOR, managed ODC, or own entity — before kickoff. Half of our recent data searches that ran over budget did so because the JD was specifying the wrong role. Get the role definition straight before opening the search.

If a strong candidate is mid-process and you’re starting to feel a counter-offer coming, our counter-offers guide covers the playbook for keeping the search alive.

FAQ

Is the data and ML talent pool in Belarus still active in 2026?

Yes. Data engineering specifically remains one of the deeper specialties in the local market, anchored by the agency ecosystem and the gaming/fintech analytics legacy. Mid-to-senior generalist DE is where the market is deepest. Senior AI engineering is scarcer, but it exists.

What’s the realistic salary for a senior data engineer vs. a senior data scientist?

Senior DE typically lands in the €4,000–6,500 EUR gross monthly range for a 5–8 year candidate with strong AWS or GCP, Snowflake or BigQuery, Airflow, and dbt. Senior DS sits in the €4,000–6,800 range with ML Engineer at the top. Senior AI engineer with real LLM production experience can land at the top of the Lead / Principal band, €6,800–9,500+. Numbers shift quarter to quarter; verify current data before publishing an offer.

Should we hire a data engineer or a data scientist first?

Almost always a data engineer first. Without clean pipelines and a working warehouse, a data scientist will spend 60–80% of their time fixing data instead of doing analysis. The exception is at very early stage when no real data infrastructure is needed yet and you’re hiring for product analytics intuition — in which case a strong analytics-leaning DS or a senior analyst makes more sense.

Is AI Engineer a separate role from Data Scientist in 2026?

In how the market behaves: yes. In how every employer titles roles: not consistently. The work is different — AI engineers spend most of their time on retrieval, evaluation, prompt and context engineering, and production infrastructure for LLM-driven systems. Classical DS work — modeling, experimentation, causal inference — is still its own discipline, just with shrinking headcount at the senior level. If you’re hiring for LLM and RAG work, write the JD as AI Engineer and budget accordingly.

How do we test SQL skills realistically during an interview?

A 30-minute live exercise with a realistic schema and a few non-trivial questions: window functions, a moderate join with aggregation, a question that requires recognizing a common SQL anti-pattern. Generic LeetCode-style problems are weaker signal. The strongest test is one where the candidate has to ask clarifying questions about the data before writing the query — that filters for someone who actually thinks about the business problem.

What’s the difference between a Data Scientist and an ML Engineer in the Belarusian market?

DS leans toward modeling, statistics, and experimentation, with deliverables that are often analyses, reports, or model prototypes. ML Engineer takes models to production, owns the infrastructure they run on, and is increasingly expected to handle LLM workflows. The skill overlap is real (both write Python and know ML frameworks), but the day-to-day work and the appropriate interview design are different.

Are remote-only data roles normal in Belarus?

Yes. Belarus has been remote-first in IT for years, and data roles run remote by default. A physical office is a nice-to-have, not a requirement. Don’t make hybrid a deal-breaker in screening; you’ll lose strong candidates over a preference rather than a need.

Is the Kaggle experience meaningful in 2026?

For data scientists specifically — yes, when it’s recent and reaches medal-level competitions. It correlates with practical modeling skill and a working comfort with messy data. For data engineers and AI engineers, Kaggle is mostly noise; their relevant public-portfolio signals are GitHub contributions and Hugging Face profiles respectively. Don’t require Kaggle for any role except DS, and even then treat it as a strong signal rather than a requirement.

Ready to start?

Tell us which role you’re actually hiring for — Data Engineer, Analytics Engineer, Data Scientist, or ML / AI Engineer — and we’ll come back with a sourcing plan, current salary range, and realistic timeline tailored to your stack and seniority. Get in touch and we’ll have it to you within 24 hours.