In 8 weeks you will design, build and deploy a real enterprise AI agent — one that remembers users, reads documents, talks to external systems and runs 24/7 on AWS. A production system, live at a real URL.
By the end of week 7 your agent is in production. Week 8 you build a second one, from scratch, for a real business case.
Part of Barcelona Code School's AI engineering track — practical skills, real tools, production code from day one.
Every week you work on the same agent — a smart assistant for a business. It starts simple and gets harder.
Week 1 — the agent answers questions about the business, searches the web for anything it doesn't know, and runs on three different LLMs so you can see the difference in quality and cost.
Week 2 — it gains memory. It knows who each client is, their preferences, their history, their interactions. Fifty simultaneous users, zero data crossover.
Week 3 — it reads the business's actual documents. Price lists, employee bios, internal policies, client contracts. It cites its sources and never makes up an answer. It also accepts a photo of a product or issue and gives a first-pass assessment.
Week 4 — you rebuild it on LangGraph with a proper state machine and connect LangSmith. Now you can trace every decision, catch every hallucination, and escalate to a human administrator when confidence is low.
Week 5 — one agent becomes three: an intake specialist, a triage agent, and a booking agent. They coordinate, pass context between them, and pause for human approval before confirming anything critical.
Week 6 — it connects to real systems. A client describes a request, attaches a photo, and without any human involvement the agent creates a lead in CRM, checks Google Calendar, books the appointment and sends a WhatsApp confirmation.
Week 7 — Docker, AWS, HTTPS, monitoring with Telegram alerts, prompt injection protection, model routing, semantic caching. Cost per request drops 40–60%. The agent is in production.
Week 8 — you design and build a second enterprise agent entirely on your own, for a real business case you choose, and present it in front of the group.
A typical student is a developer who wants to specialise in AI engineering to stay relevant for the "AI-first" and "AI-agent" engineer roles.
Or someone from a technical background who has been watching the AI agent space and decided it is time to get in properly.
You graduate with two working enterprise agents in production and a portfolio that shows real engineering decisions.
The AI agent engineering market is moving fast. Companies are hiring people who can build these systems right now, not in two years when university programmes catch up.
Become prepared for positions like AI agent engineer AI integration developer, Conversational AI engineer, Automation engineer.
You will also leave knowing how to cost and evaluate the agents you build — a practical skill that separates engineers who ship production systems from those who only demo them.
Paid at registration.
Instalment payments are available at the checkout by 
8 weeks, 40 hours/week, 320 hours total
Monday–Friday, 9:00–18:00
On campus in Barcelona or remote live
Small cohorts — spots fill fast
8 weeks, 40 hours/week, 320 hours total
Monday–Friday, 9:00–18:00
On campus in Barcelona or remote live
Small cohorts — spots fill fast
Basic is enough — variables, functions, libraries, HTTP requests. You do not need to have built anything with AI before. The first week covers everything you need to start working with LLMs.
Yes. Both tracks attend the same live sessions simultaneously, see the same instructor screen, and do the same hands-on builds in real time. Same sessions, same price.
Any Mac or Windows laptop made in the last five years. The heavy compute runs on cloud services — your machine just needs to run Cursor and a browser.
Yes. During the course you will use paid APIs — OpenAI, Anthropic, Google — and AWS services. Expect to budget roughly 50–100€ over the 8 weeks for API and cloud costs. We teach cost management as part of the curriculum so you keep this under control.
It is yours. It runs on your own AWS account. There are no ongoing fees to Barcelona Code School.
Recordings are available. That said, this is a hands-on bootcamp — the practical work done in sessions is where the real learning happens. We recommend attending live wherever possible.
English.
Yes. Before the first session we will send you setup instructions (accounts to create, tools to install) and a short Python refresh to complete. It takes 2–3 hours and makes sure everyone starts day one at the same level.
Yes. An industry-recognised Barcelona Code School certificate upon completion of the bootcamp and the final project.
