AI Fundamentals
Masterclasses
Structured learning programs covering machine learning, neural networks, and practical AI implementation — taught by practitioners who use these tools daily.
Six programs, one clear direction
Each program focuses on a specific layer of AI — from understanding how models learn to deploying them in real products. You pick the depth that matches where you actually are, not where you wish you were.
AI Foundations
A 4-week program covering how machine learning models actually work — gradient descent, loss functions, and the math behind predictions. No shortcuts, no hand-waving.
BeginnerNeural Network Design
Covers architecture selection across 6 network types — CNNs, RNNs, transformers, and their hybrids. You build 3 working models during the course, not just diagrams.
IntermediateData Preparation Lab
Working with messy real-world datasets — 80% of AI work is here. The program covers cleaning, labeling pipelines, and feature engineering across tabular and image data.
PracticalPrompt Engineering
A focused 2-week intensive on working with large language models. Covers chain-of-thought prompting, few-shot examples, and systematic evaluation of model outputs.
AppliedModel Deployment
From notebook to production — covers containerization, REST API design, and monitoring. You deploy a working inference endpoint by week 3 of the 5-week program.
AdvancedAI Ethics and Audit
Bias detection, fairness metrics, and documentation practices for AI systems. Covers the EU AI Act framework and how it applies to systems built in 2024 and beyond.
SpecialistHow a session actually runs
Each masterclass runs across 5 live sessions of 90 minutes each, with recorded replays available within 24 hours. Groups stay under 18 participants so questions get real answers, not generic ones.
The structure below applies to every program — the content changes, the format stays consistent so you always know what to expect.
Concept walkthrough
The instructor explains the core idea with a concrete example — not a textbook definition. Usually 25–30 minutes with live questions.
Guided implementation
You follow along in a shared notebook environment. Code runs in your browser — no local setup needed for the first 3 sessions of any program.
Independent task
A small assignment on a different dataset than the demo. Takes 45–60 minutes. Instructor reviews submissions before the next session.
Feedback and iteration
The next session opens with a review of common mistakes from the task. Specific, named feedback — not generic "good effort" responses.
The people who teach here
Instructors at Lemdranok come from engineering and research backgrounds — they are not professional educators who learned AI to teach it. Most still work on active projects.
The Neural Network Design program was harder than I expected — in a good way. By session 4 I was debugging a transformer architecture I had written myself. That does not happen in most courses.
— Dmytro Savchenko, software engineer, KyivI took the Data Preparation Lab after two years of working with tabular data and still learned 9 or 10 things I had been doing wrong. The section on label noise alone was worth the time.
— Iryna Kovalenko, data analyst, Lviv