Crash Course: Building a RAG system
Mentor - Natalia Manakova, Senior Data Scientist, AI consultant at SoftServe, PhD
Updated and expanded course on building a RAG system: in the attached session, you will learn more about how to assemble a RAG completely locally, without getting your private documents on the network.
Most AI tools work well with general knowledge, but struggle with real documents: internal knowledge bases, PDFs, notes, customer data, etc. This is where RAG (Retrieval-Augmented Generation) comes in, an approach that allows you to combine searching your data with the capabilities of LLM.
Also, RAG systems are actively used in products: from chat bots to corporate knowledge systems, which makes them very relevant for the modern market.
In this crash course, you will step-by-step put together your own RAG system that will answer questions based on your documents and add references to sources.
- In the process of work, you will understand the key components of modern AI solutions:
- how does semantic search work
- what are embeddings and how are they used
- how vector databases are arranged
- how to combine retrieval and generation into one system
The format combines short theory, live coding and practical examples. You will not only understand how RAG works, but also be able to independently implement the basic version and adapt it to your tasks.
As a result, you will receive an example of a ready-made system and a clear understanding of how to apply these approaches in practice for real time savings and increased team efficiency.
Audience
- Data Scientists / ML Engineers (classic ML → GenAI)
Experienced in classic ML who want to move to GenAI and gain hands-on experience with LLM, embeddings, and RAG. - Data / Business Analysts
Analysts who work with large amounts of text data and want to automate information search and analysis using AI. - AI / GenAI enthusiasts (with basic knowledge)
Experts familiar with ChatGPT or OpenAI API who want to move from prompt experiments to building full-fledged AI systems. - Tech Leads / Architects
Technical leaders who evaluate and implement AI solutions and want to gain a deeper understanding of the RAG architecture and its application in production.
Format:
Date and time: August 27, 27, 29, online, start at 18:30 (Kyiv time, GMT+2) on weekdays, at 10:00 on Saturday.
Duration of each session: ~2.5 hours
Platform: Zoom. You will receive a link to the broadcast the day before the workshop, and it will also be available on this landing page.
Recommended toolset and subscriptions:
Event and presentation language: Ukrainian
Program
- Basics: what is RAG and why is it needed
- Architecture: how the RAG system (retrieval + generation) works
- Data: preparation of documents (PDF / text)
- Data preparation: chunking and embeddings for effective search
- Infrastructure: vector databases
- Integration: connection to LLM
- Implementation: pipeline construction
- Practice 1: Demonstration and testing of the system on OpenAI models + Telegram chatbot.
- Practice 2: Analysis of costs and evaluation of the quality of answers.
- Practice 3: Assemble the pipeline entirely locally. The nuances of working with local LLMs.
Mentor: Natalia Manakova, Senior Data Scientist, SoftServe
— Has experience in developing AI solutions using LLM and RAG
— The portfolio includes production-ready GenAi/RAG/Agentic systems
— PhD in Applied Mathematics and Advanced Information Technologies
— Speaker at conferences and mentor in WomanWhoCode programs
— Conducts research on AI in the company's internal lab
Event price
Access to online course sessions
Access to videos and materials on the educational platform
Presentations
Access to the chat, where the speaker will answer questions
Certificate of the participant (subject to completion of homework)
Free month at Fwdays Club
10% discount on participation in Fwdays conferences
Available payment in installments from Monobank and purchase in installments from PrivatbankIf after the first lesson you realize that the expectations do not match, we will refund the money in full.