/workshops/tech-teamsPe română

AI workshop for developers and engineering teams

Vlad Tudor runs hands-on AI workshops for developers and engineering teams in Romania on working with coding agents: writing the spec first, breaking work into tasks an agent can finish, and reviewing what it produces. He is an AI engineer with eight years of production work and founded Sapio AI, a studio whose engineers build this way every day.

Reply within 24 hours on working days

An engineering office at night, code on the monitors and a system diagram on the glass
Format
Full day, 9:00–17:00 · 1 or 2 days
Group
Up to 20 participants
Language
Romanian or English
Materials
Built on your own cases

What I hear from engineering teams.

“It writes code fast. Then I spend the afternoon finding out what it broke.”

“We have the licences. Two people use them well and nobody knows how.”

“I'm not letting an agent near our codebase.”

The tool is not the hard part. The method is.

You don't get a sceptical developer to use AI by mandate. You get there by starting on the side project they never had time for, away from the fragile production code, until they have a feel for how to steer it.

The rest is method, and the method is old: define the work precisely before anyone starts, break it into pieces with clear inputs and outputs, and don't accept anything that hasn't been reviewed and tested. What changes is that the one doing the typing is no longer the scarce resource. Defining the work is.

How to get a sceptical developer to use AI: start with the side project he never had time for.REWIRED · 21:15–21:57 · In English

From the REWIRED podcast.

One day, 9:00–17:00.

Morning · 9:00–12:30Method
  • What a coding agent actually does in a loop, and where it goes off course.
  • Spec first: writing the task, the constraints and the definition of done before the first prompt.
  • Decomposition: splitting a feature into tasks an agent can finish and a human can review.
  • Project context: the conventions file the agent reads, so it stops guessing your architecture.
  • Guardrails: permissions, secrets, sandboxes, and what never runs unattended.
12:30–13:30 · Lunch
Afternoon · 13:30–17:00Hands-on
  • Exercise 1: a side project each, from spec to working version, with the agent doing the typing.
  • Exercise 2: one real ticket from your backlog, run in a sandboxed copy of your repo: plan, execute, review the diff.
  • Exercise 3: tests and a small evaluation written first, used as the gate for the agent's work.
  • Round-up: a team working agreement on which tasks go to agents and how the output gets reviewed.

On Monday, developers can:

  • write a spec an agent can execute and a colleague can check;
  • split a ticket into tasks the agent finishes without drifting;
  • review agent output the way they review a colleague's pull request;
  • keep secrets and production systems out of the agent's reach.

Rooms and work behind it.

  • Eight years of production AI

    Building production AI: agents, retrieval systems, speech and vision models.

  • Sapio AI

    Founder of a studio of AI-first engineers who build with agents on client projects that end in production.

  • Romanian AI Alliance

    A workshop for developers and engineers with the Romanian AI Alliance.

  • Politehnica University of Bucharest

    Four years teaching engineering students.

What's included.

  • A preparation call with the tech lead about your stack, tools and codebase.

  • Exercises built on your stack and a sample of your own tickets, in Romanian or English.

  • The day itself, on site, with the team's own machines and tools.

  • A short check-in after the workshop, agreed in the proposal, on how the working agreement holds up in practice.

One day (9:00–17:00) or two days, the second on a real piece of your product. Up to 20 participants. Romanian or English. On site. Works with the coding tools your team already uses.

Weighing a specific AI decision rather than training a team? AI consulting

Questions that come up.

How much does it cost?

It depends on the format and the size of the group. After a short call you get a written proposal before anything is agreed.

Online or on site?

On site, with people on their own machines, because most of the learning happens when I can look at someone's screen. Online delivery is possible on request, discussed case by case.

In Romanian or English?

Either. Code and tools stay in English anyway; the explanation can be in Romanian.

How many developers?

Up to 20, and fewer is better for the afternoon. A larger team can be split into groups on different days.

Does it cover Article 4 of the AI Act?

Article 4 asks organisations that use AI systems to take measures to support AI literacy among the people using them on their behalf, and developers using coding agents are included. A workshop is one such measure, and the agenda and materials remain as a record. The Act asks for effort, not a certificate. I don't give compliance guarantees; your legal adviser decides whether your measures are enough.

Tell me who would be in the room.

Tell me your stack, the tools the team has and what hasn't worked so far. I'll build the day on your own backlog.

Reply within 24 hours on working days

Group size

How many people?