Agentic AI course, live online

Agentic AI course for developers: build AI agents that work in production

A 10-week live program where you build AI agents in Python: tool calling, memory, MCP servers, multi-agent systems, RAG, guardrails and deployment. Projects are checked by autotests, and an instructor reviews your GitHub projects. Certificate on completion. Next cohort starts November 23, 2026.

10 weeks, live$2,299 with code AUTUMN5 GitHub projectsCertificateWith JetBrains Academy
See dates and enrollPrice and payment plans

What you build, week by week

Weeks 1 to 2
LLM pipelines and your first agent tools
OpenAI API, prompting, LangChain basics, a CLI chat and MCP servers that give the model tools.
Weeks 3 to 4
Agents and multi-agent systems
First an agent from scratch in plain Python, without frameworks. Then tool calling, memory and reasoning loops. Projects: a GitHub pull request review agent and a multi-agent backend on FastAPI.
Weeks 5 to 7
Agents with knowledge: vector databases and RAG
Embeddings and semantic search with Qdrant, chunking, retrieval, query rewriting, reranking, HyDE.
Weeks 8 to 9
Guardrails, evaluation and cost control
Evaluation with Langfuse and Ragas, NeMo guardrails, cost limits, caching with Redis. The part most agent courses skip.
Week 10
Deploy the agent
FastAPI service, Docker, secrets and configs, deployment to AWS with GitHub Actions.
Demo Day
Graduation project
An agent for a real problem, with architecture, evaluation and trade-offs, presented to the cohort and instructors.

Is this a separate course?

No. This page describes the agent track of the Hyperskill AI Engineer Bootcamp. Agents are the core of the program, and the rest of it, RAG, evaluation, monitoring and deployment, is what makes an agent usable in production rather than in a notebook. If you want agents only, the bootcamp is still the shortest path: the agent modules alone would leave you without the parts that break first in production. Full program, dates and enrollment.

Who it is for

Software developers
You code in any language and know basic Python. No machine learning background needed, the course starts from the API, not from the math.
Tech leads and architects
You decide where agents fit in your system and need to judge trade-offs: latency, cost, reliability, security.
Engineering teams
The same program runs as a private cohort for 4 to 20 engineers, on your stack and your use cases. Agentic AI training for teams.

About the certificate

You get a Hyperskill certificate of completion issued with JetBrains Academy. If you are searching for an agentic AI certification: there is no industry-standard one yet, and no vendor exam makes you an agent engineer. What hiring managers check is the code. By the end of the course you have 5 public GitHub repositories and a graduation project, and that is the proof we build the program around.

Tools and frameworks

PythonOpenAI APILangChainLlamaIndexMCPFastAPIQdrantLangfuseRagasNeMo GuardrailsRedisDockerAWSGitHub Actions

Instructors

Nikolay Vyahhi
Founder of Hyperskill, AI educator, MIT lecturer. Program author.
Nikalina Ogorodova
AI/ML engineer, instructor on the program.
Ruslan Davletshin and Ivan Rodin
CTO and AI researcher. Open hours, evaluation, deployment and production trade-offs.

Questions about the agentic AI course

Do I need a machine learning background?
No. You need programming experience and basic Python. The course works with LLM APIs and frameworks, not with model training.
How much time does it take?
10 to 12 hours a week for 10 weeks: 1 to 2 hours of live sessions, 5 to 7 hours of projects, 1 to 2 hours of self-paced lessons. Sessions are recorded.
How much does it cost?
$2,499, or $2,299 with the code AUTUMN for the November 23 cohort. A 3-payment plan and Klarna for US residents are available. Full refund until the end of the first week.
Which agent frameworks do you teach?
LangChain and LlamaIndex for pipelines and agents, MCP for tools, FastAPI for serving. The point is the patterns: tool calling, memory, orchestration, evaluation, which transfer to any framework.
How is it different from free agent tutorials?
Tutorials stop at a working demo. Here projects are checked by autotests, an instructor reviews your GitHub projects, you deploy to AWS, add evaluation and guardrails, and present a graduation project at Demo Day.
Can my team take it together?
Yes, as a private cohort for 4 to 20 engineers with the curriculum adjusted to your stack. From $1,500 per engineer.
See dates and enrollRead graduate reviews