PopiAI
Download the dossier

Infrastructure & DevOps

Harness Engineering training

Advanced agent control, feedback loops and production patterns, for infra and DevOps.

30+ teams trained · 9.6/10 · Certified instructors
Request a proposal →Download the dossier

Your agents work in the demo. What about production?

Getting an agent to answer well once is easy. Making it reliable, auditable and maintainable in production is a different problem — the one harness engineering solves.

  • For infra and DevOps.
  • Teams that already run agents but with no quality control, guardrails or traceability.
  • Organizations that need their agents to run safely and auditably in CI.

What you'll learn

Modules, duration and format

  • Fundamentals and specifications
  • Automations, deployments and a simple harness
  • Advanced harness engineering
Duration
1-2 days
Group
Up to 15 people
Format
In-person, in-company (anywhere in Spain) or remote

Methodology

Hands-on, on real cases

Generic courses that scratch the surface, videos nobody finishes and zero impact on real work. Meanwhile, teams that truly know how to use AI do in hours what takes others weeks.

  • Real diagnosis

    We map the team, the stack and the processes, and pick the use cases that actually move the needle.

  • Hands-on, on your stack

    Sessions building agents and automations on your real tools, not toy examples.

  • You leave applying it

    The team walks away with judgement and things working, ready to maintain them and build more on their own.

Who teaches it

Practitioners who build with AI every day

  • Fran Sánchez

    Fran Sánchez

    Lead trainer · Builders

    “If I don't use it every day, I don't teach it.”

    15 years at the intersection of product and artificial intelligence — from the first neural network running on a phone keyboard to ad products generating billions a year at Meta.

    • Meta
    • Microsoft
    • SwiftKey
  • Luis Miguel Gil

    Luis Miguel Gil

    Harness · AI infrastructure

    “If it doesn't scale or can't be maintained, it's not worth it.”

    10+ years as an SRE/DevOps and backend engineer at Voicemod, Landbot and other startups — now an independent AI & DevOps consultant.

    • Leads the technical side of the Claude Code programs — harness engineering and autonomous agents in CI (PR reviews, overnight repository audits) that he runs daily on his own projects.
    • Trains teams at startups and established companies in agentic development: what he teaches in the classroom is what he ships in his repos that same week.
    • CTO of Roostwood: an LLM copilot embedded in 3D design software — the model emits structured changes and the geometry regenerates deterministically. AI that reaches production and stays there.
    • AI and backend engineer at Button: backend and cloud infrastructure from scratch to Google Play and the App Store, with ML in production on small models — chosen for cost, not for hype.
    • Previously an SRE and DevOps engineer at Voicemod and Landbot: infrastructure, CI/CD and observability for products with millions of users. He knows where the money hides in a cloud bill.
    • Landbot
    • Roostwood

Real results

Not theory.

  • “It truly changes your perspective on how software is built today: you go from writing code to leading your own team of agents — PM, architect, dev, QA. Invest real time in a solid spec and development comes out a thousand times faster and more structured.”

    Andrés GuerraComputer Engineer, Beemining Consulting

Price

Tailored proposal

Contact

Request a proposal

Tell us about your team and we'll send you a tailored proposal.

What's in the proposal?

  • Priority use cases for your team
  • Recommended program and format
  • Calendar and support plan

Common questions

FAQ

What is harness engineering?

It's the advanced control of AI agents: feedback loops, guardrails, traceability and production patterns, so an agent that works in a demo also works reliably in your CI/CD.

Is it in-person and in-company?

Yes. We train in person at your offices, in small groups and hands-on, and we travel anywhere in Spain. We also deliver remotely in Chile, and very soon in Colombia, Venezuela and the UAE.

How many people can take part?

We work in groups of up to 15 people, to keep the attention direct and make sure everyone builds.

Do you adapt it to our industry and tools?

Yes. We design the program on your real cases, data and tools — nothing pre-recorded or generic.

How does the proposal work?

After a short call to understand your team, we send you a tailored proposal with cases, itinerary and a support plan.

Is it a standalone module or part of Claude Code?

It's one of the Claude Code for development teams tracks — you can hire it on its own or as a follow-up to Claude Code Standard/Custom.

Other programs

Keep exploring

  • Claude Code training for development teams

    From coding to delegating to agents, with a production-grade method. For developers, tech leads, SRE and DevOps.

    See the program →
See all programs →Training for companies →