Crash Course: Deep Dive into LLM APIs
Mentor - Oleksandr Krakovetsky, AI Expert | CEO at DevRain | CTO at DonorUA | Author of books on generative artificial intelligence
Buy tickets for the next conference Fwdays AI Summit!
This course is a practical and systematic guide to working with large language models for engineers, architects, and technical leaders who want to not just use LLMs, but understand how they work and how to build solutions for their projects.
You will understand the internal architecture of transformers, key properties and limitations of modern language models, learn how to correctly configure them via API, and work with typical problems - in particular, hallucinations, contextual limitations, and unstable response quality.
Special attention is paid to the transition from Prompt Engineering to Context Engineering: how to build systems where the model works stably within the given rules, knowledge, and tools. You will learn how to use LLM as a backend component, build RAG solutions, connect external tools, work with structured responses, and optimize token usage in real production scenarios.
The course combines theory with practical demos so that you can immediately apply the approaches in your own products, services, or internal platforms.
You will understand in detail:
- how modern language models work "under the hood"
- what mechanisms determine their behavior
- how to apply this knowledge to create reliable, manageable, and scalable solutions
A separate block is dedicated to working with benchmarks:
- how to correctly evaluate models
- what characteristics have practical significance
- how to interpret test results on instruction-following, working with long-term context, and other metrics
After the course, you will be able to test models for relevance to specific tasks and make technical decisions based on metrics, not intuition.
Who will be interested?
- Software developers
- Tech leaders
- Architects
- Anyone who wants to understand how language models work under the hood
Format:
Date and time: February 24&26, online, start at 18:30 (Kyiv time, GMT+2)
Duration of each session: ~2.5 hours
Theory and practice: 50/50, before the start we will send you a selection of materials to familiarize yourself with key concepts. There will be homework to consolidate the material and understand the concepts.
Platform: Zoom. You will receive a link to the broadcast on the eve of the workshop, and it will also be available on this landing page. In addition, all materials and lecture recordings will be available on the learning platform.
Recommended toolset and subscriptions:
- OpenAI API Playground (Anthropic API Playground, Google AI Studio Playground are also ok)
- Ollama
- Postman or other API-manager
Language of the event and presentations: Ukrainian
Program
Day 1
- Transformer architecture
- Key concepts, types, characteristics, and functionality of large language models
- API parameters for configuring the operation of large language models
- API evolution
- Metrics for assessing the quality of models
- Key characteristics of large language models (multimodality, hallucination, reasoning, non-determinism)
- From Prompt Engineering to Context Engineering
Day 2
- Structured responses
- Token optimization
- Building RAG solutions, typical problems
- Fine tuning
- Evals
- Instrumentalization: function calls, using external tools
- Benchmarks for model quality assessment ("instruction following", working with long context, AGI, etc.)
Mentor: Oleksandr Krakovetsky
— Co-founder and CEO of the Ukrainian IT company DevRain
— Co-founder and technical director of DonorUA - an intelligent blood donor recruitment system.
— Candidate of Technical Sciences in Information Technology.
—Microsoft Regional Director, Microsoft Artificial Intelligence Most Valuable Professional.
—Microsoft Certified: Azure Data Science Associate