🤖 ML Engineer Roadmap 2026

Roadmap to Senior ML Engineer in the Russian market

Русская версия · English version

---

📍 About the roadmap

Materials for getting an ML Engineer offer.

---

🗺️ Roadmap

flowchart TD
    A[🚀 Start] --> B[1️⃣ Foundations<br/>Python · Math · CS]
    B --> C[2️⃣ Data<br/>SQL · Pandas · Spark]
    C --> D[3️⃣ Classical ML<br/>sklearn · CatBoost · XGBoost]
    D --> E[4️⃣ Deep Learning<br/>PyTorch · CV · NLP]
    E --> F[5️⃣ Modern AI<br/>Transformers · LLM · RAG]
    D --> G[6️⃣ MLOps<br/>Docker · K8s · MLflow · CI/CD]
    F --> G
    G --> H[7️⃣ System Design<br/>for ML systems]
    H --> I[💼 Junior ML Engineer<br/>120–200k ₽]
    I --> J[💼 Middle ML Engineer<br/>200–350k ₽]
    J --> K[💼 Senior / Lead<br/>350–800k+ ₽]

    style A fill:#4A90E2,stroke:#fff,color:#fff
    style I fill:#27AE60,stroke:#fff,color:#fff
    style J fill:#16A085,stroke:#fff,color:#fff
    style K fill:#8E44AD,stroke:#fff,color:#fff

Loading

---

📚 Stages

#	Section	Duration	EN	RU
1	🐍 Foundations — Python, algorithms, CS	1–2 months	en	ru
2	📐 Math for ML	in parallel	en	ru
3	🗄️ Data Engineering for ML	1–2 months	en	ru
4	🤖 Classical ML	2–3 months	en	ru
5	🧠 Deep Learning	2–3 months	en	ru
6	💬 NLP, CV, LLM	2–3 months	en	ru
7	⚙️ MLOps & Production	2 months	en	ru
8	🏗️ ML System Design	1 month	en	ru
9	💼 Career in the Russian market	continuous	en	ru
10	🎯 Pet projects and portfolio	continuous	en	ru
11	📖 Resources	—	en	ru

---

🎯 Approximate plan

gantt
    title Approximate 12-month plan
    dateFormat  YYYY-MM-DD
    axisFormat  %m.%y

    section Foundations
    Python advanced + algorithms   :a1, 2026-05-01, 45d
    Math review + statistics       :a2, 2026-05-01, 60d

    section Data
    Advanced SQL + Pandas          :b1, after a1, 30d
    Spark / Airflow basics         :b2, after b1, 30d

    section ML
    Classical ML + Kaggle competition :c1, after b1, 75d
    Deep Learning with PyTorch        :c2, after c1, 75d
    NLP / LLM / RAG                   :c3, after c2, 60d

    section MLOps
    Docker, FastAPI, MLflow, CI/CD :d1, after c2, 45d
    K8s + serving, Triton/vLLM     :d2, after d1, 30d

    section Career
    Pet projects + GitHub          :e1, 2026-06-01, 300d
    Internship / offer             :e2, 2026-11-01, 180d

Loading

---

💰 Salary benchmarks, Russia 2026

Grade	Experience	Moscow, gross / month	Regions
👶 Intern / Trainee	0	60–120k ₽	40–80k ₽
🚀 Junior	0–1 year	120–200k ₽	90–150k ₽
⚡ Middle	2–3 years	200–350k ₽	150–280k ₽
🔥 Senior	4–6 years	350–600k ₽	250–450k ₽
👑 Lead / Staff	6+ years	600k–1M+ ₽	400–700k ₽

Sources: hh.ru, getmatch, Habr Career, Habr salary reports, ODS chats. Numbers depend on company, grade, and stack; LLM/CV roles usually pay above average.

---

🏢 Who hires ML Engineers in Russia

🥇 Tier-1	🥈 Tier-2	🥉 Tier-3
Yandex (Search, Alice, Shedevrum)	Avito	Wildberries
Sber / SberAI (GigaChat, Kandinsky)	Ozon	X5 Tech
T-Bank	MTS AI / MWS	Alfa-Bank
VK (VK Tech, Marusia)	Kaspersky	Gazprombank Tech
	Skoltech / AIRI	VTB, Sovcombank

---

✅ Interview readiness checklist

• Python: OOP, async, typing, tests (pytest)
• Algorithms: ~150 LeetCode Easy/Medium problems
• SQL: window functions, optimization, EXPLAIN
• Math: linear algebra, calculus, probability, statistics
• Classical ML: explain bias/variance, regularization, boosting internals
• Deep Learning: implemented backprop, CNN, Transformer by hand
• PyTorch: can write a train loop without copy-paste
• LLM: attention, KV-cache, RAG, fine-tuning (LoRA)
• MLOps: deployed a model with Docker + FastAPI + monitoring
• System Design: studied 3+ cases: recommendations, search, fraud, NLP service
• Portfolio: 2–3 strong pet projects on GitHub
• Kaggle: 1+ competition with a Bronze+ medal
• Resume on hh + getmatch + Habr Career

---

🧭 How to use this repository

1. Go through sections in order, but study math and algorithms in parallel.
2. Reinforce every topic with a project; see 10_projects.md.
3. Every 2 weeks, review progress and update PROGRESS.md.
4. Most resources are Russian-language and practical for the Russian market.

---

⭐ If this roadmap is useful, give it a star

Made for learning and portfolio building · 2026