Basic knowledge of programming in any language. For the practical exercises, being able to “read and understand” Python code is helpful – and fully sufficient. Previous experience with AI or DevOps is not required; we guide you through everything from the basics.
The course is primarily aimed at developers and software specialists who want to introduce artificial intelligence and autonomous agents into their profession, as well as those who want to move into the emerging roles of AI programmers and architects.
Want to build autonomous agents with memory, tools and data instead of isolated chatbots.
Design systems and need to understand the architecture of multi-agent and RAG solutions in production.
Want to master running LLMs, fine-tuning and deploying models on the NVIDIA stack.
Need to run AI applications on Kubernetes with monitoring and scaling.
The course is taught online, so you can join from anywhere. Lessons combine explanation, demonstrations and practical exercises, and video recordings of all lessons are available for you to return to at any time.
All lessons are available in the e-learning platform for unlimited playback.
Practical exercises on our GPU instances – no expensive hardware needed.
You try out new skills right away on real-world tasks.
Between lessons (or between weekends) you can work on assignments with feedback from lecturers.
A space for questions, sharing solutions and communicating with other participants.
Real hardware for running LLMs and agentic systems within reach.
We offer the course in two formats – the specific days and times are always shown on the card of the chosen date.
In this block we work with the LLM models themselves, including tuning and running them with NVIDIA tools. It is an essential foundation for anyone who wants to take their skills to a professional level and obtain official NVIDIA certification.
In an accessible way we travel from a single neuron through multilayer networks, recurrent neurons and embedding layers to the Transformer architecture. You will understand how modern language models really work.
From theory to practice – you build your first neural network. You learn to design the architecture, prepare data, train the network and evaluate its results. You gain the practical foundation that all modern AI models are built on.
We turn theory into skill. Guided by the lecturer, you build and train your own neural network, try tuning its parameters and see how individual decisions affect the final accuracy.
You learn to get maximum performance out of models – optimisation with TensorRT-LLM and practical fine-tuning and training in NVIDIA NeMo. You gain clarity on a key decision: when fine-tuning a model (NeMo) pays off and when RAG is the better choice.
You master deploying models to the real world as modern microservices. We work with Triton Inference Server and the latest NVIDIA NIM and compare NIM vs. Triton vs. self-hosting so you can choose the ideal solution for any environment.
You find out what modern AI really runs on. We cover GPUs and their memory, the role of VRAM in inference and training, model quantisation, and you learn to estimate what hardware you need – from running smaller models locally to choosing a GPU in the cloud.
This block takes us through the dynamic world of AI agents. We explain how an agent works and get hands-on building cooperating systems – from your own knowledge base through RAG to the architecture of complex distributed agents.
You master techniques for safe and effective control of models – defence against prompt injection, output validation and policy enforcement. We also show how to involve a human via Human-in-the-loop.
You understand how real AI agents are built – how they think, plan their steps and use tools. You learn to compose logic into clear, reusable components that form the basis of real AI applications.
You get hands-on with LangChain. Guided by the lecturer, you build your own agent, connect it to tools and test its behaviour on real tasks.
You build agents that remember context and can cooperate effectively. You learn to create systems with different roles that hand over tasks and solve complex problems together – the foundation of next-generation autonomous systems.
We turn theory into skill. Guided by the lecturer, you create and configure your own system of cooperating agents and verify their capabilities in a real environment.
We give agents the ability to remember the past. We explore MCP for long-term memory, vector databases for semantic search, SQL/NoSQL for chat history and the use of graph databases as a knowledge graph.
You teach agents to retrieve and store new data. We show how to connect models to documents and databases using RAG, how vector databases work and how to implement advanced semantic search on top of them.
You create your own knowledge base, connect it to agentic systems and teach the AI to work effectively with your specific data.
You learn to build agents that can be trusted. We discuss the security risks of agentic systems, protecting sensitive data, guardrails and limiting agent permissions, as well as the ethical questions of deploying AI – from transparency of decisions to responsibility for outputs.
Become an architect of robust AI solutions. We explore the trade-offs between speed, accuracy and price (latency vs. accuracy vs. cost), caching strategies, retry and fallback mechanisms, and multi-agent and event-driven architectures for scalable processes.
You learn to keep your agents under control even after deployment. In LangSmith you try step-by-step tracing of agent runs, prompt tuning, evaluating answer quality and tracking cost and latency – everything you need to run AI applications reliably in production.
An integral part of building agents is running them reliably – the standard (also required for NVIDIA certification) is Kubernetes. On dedicated GPU instances in our unique lab we get hands-on experience running agents and building RAG systems in a real environment and prepare for the certification exam.
You get to know the basic building blocks of container orchestration – Deployment, Pod, PVC and SVC. You learn to work with the Helm package manager and understand the role of the Registry for managing images.
We connect to our GPU lab. Guided by the lecturer, everyone tries the basics of configuring a Kubernetes cluster and real deployment of applications, including the agents themselves.
We focus on preparing the environment for RAG both locally and in the AWS lab. You master the NVIDIA GPU CLI and Operators, understand the architecture of a RAG pipeline and learn to use it effectively in production.
You master the paths data flows through – Ingress controllers, internal communication via SVC and the differences between Load Balancer, Node Port and Cluster IP.
We work with the Milvus vector database, try reranking techniques and deploy a complete RAG system so that agents work with data accurately and quickly.
You learn to track the health and performance of agents in real time, LLM-specific metrics and automatic cluster scaling for both burst load and long-term operation.
You bring everything together into one real-world whole. In the cluster you deploy and run a complete AI application – LLM, agents and RAG together – and try scaling and monitoring it the way real production operation requires.
We go through the detailed course of the NVIDIA NCP-AAI exam and the key areas the test focuses on. Our lecturers – themselves holders of the certification – share real questions, proven tips and tricks for passing the exam on the first attempt.
After completing the course you will not only understand how agentic systems work – you will be able to build, deploy and run them in a real environment.
While others talk about AI in general, we teach you the specific tools that professionals around the world use today:
Join the elite of engineers who don’t just build chatbots – they build next-generation autonomous agentic systems.
We designed the entire syllabus to overlap precisely with the requirements of this prestigious certification. Our lecturers are certified experts who have successfully gone through the process themselves and will pass on their know-how and first-hand practical tips. At the end of the course we also include an intensive preparation module focused on exam strategy and key topics.
The NVIDIA Certified Professional – Agentic AI certification represents the absolute top tier in the development of autonomous systems. It focuses not just on generating text, but on the ability to design and deploy complex agentic workflows.
It confirms that you are not just a user of AI but its architect – able to use the NVIDIA technology stack to build production agents that are fast, secure and scalable.
The exam goes deep into practice. It tests your skills in:
Agentic AI is the next evolutionary leap in IT. By earning this professional title you join a small group of experts who can transform static software into intelligent, acting systems.
For companies, the certification is a guarantee that you can build solutions that truly replace complex manual processes and deliver real economic value. It is your stamp of belonging to the elite of modern engineering.
It doesn’t end with the course. Even after completing it, you keep plenty of benefits you can rely on:
An official Turing Institut certificate documenting your newly acquired skills.
Recordings of all lessons and study materials remain yours forever – come back to them whenever you need.
A Discord full of people like you, a newsletter with IT news and exclusive members-only events.
The world of artificial intelligence changes every day – which is why our teaching doesn’t end in the classroom. As a participant you get a supplementary library of video lessons that go even deeper:
There is currently no scheduled date for this course. Write to us or give us a call — we will let you know as soon as a new date is announced, and we are happy to arrange a date that suits you.
We have submitted this course to the Czech Labour Office programme. Once it is approved, the state can cover up to 76.73 % of the course price for you. Join the waitlist and you will be among the first to know.
How the funding worksNo. All demanding practical exercises run on our GPU instances in the cloud, to which you get access. An ordinary laptop with a browser is enough – you don't need to buy expensive hardware.
Basic knowledge of programming in any language is enough. The practical part is mostly in Python, which is today's standard for working with AI – knowing it is an advantage, not a requirement. The topics covered are not demanding in terms of programming.
Yes. The whole syllabus is designed to overlap with the requirements of the NVIDIA Certified Professional – Agentic AI certification, and we include an intensive preparation module at the end. Our lecturers hold this certification. Completing the course itself, however, is not an automatic guarantee of passing the exam.
No. The only prerequisite is basic programming. We touch on machine learning and neural networks in the course and explain everything needed there.
The course is designed so that knowledge of advanced mathematics such as linear algebra or calculus is not necessary. Simple mathematics is all we need. Our lecturers are, however, well versed in mathematical and theoretical topics and are happy to answer questions.
These two courses sit side by side in the catalogue, but they teach different skills.
This course takes ready-made language models (LLMs) as the brain of an application and teaches them to act: plan steps, reach for tools, look up information in company data (RAG), remember context and cooperate with other agents. You build agents in LangChain and LangGraph, connect them to vector databases via MCP and RAG, and finally run them in production — Kubernetes, the NVIDIA stack and preparation for the NCP-AAI certification. You try it all in a real environment on our GPU instances.
Machine Learning and Neural Networks is a course about how a model learns from data. You build and train your own models — from classic algorithms (Bayes classifier, decision trees, unsupervised learning) through multilayer networks to convolutional networks for images, generative networks (GANs), autoencoders, time-series prediction (LSTM) and language processing. You work in Python with Scikit-learn and TensorFlow / Keras. The goal is to understand the principles, evaluate model quality and know how to tune it.
The courses meet, but only at the edges. Here you take an accessible journey from the neuron to the LLM and build a simple network so it is clear what language models are built on — but that does not replace a machine-learning course. Conversely, Machine Learning ends with one lesson on agents in LangChain. It is a taste of where you can go next, not the syllabus of this course.
A simple guide: if you want to build applications that already use LLMs, connect them to tools and data and run them — stay here. If you want to understand how a model learns from data, train one yourself and be able to say whether it is good — go for the Machine Learning and Neural Networks course. They build on each other, but neither is a prerequisite for the other.
A more detailed comparison of all AI courses by goal can be found in the article Which AI course should I choose?.