Agentic AI
training

Hands-on workshops, live demos,
pipelines that run themselves overnight.

Toulouse · in-house · remote

The concrete piece

AI agents in your pipeline

A pi.dev agent (the open-source agent tooling I use daily) deployed in your environment, wired to your data, running on its own at night.

What I set up at your place

  • The pi.dev harness deployed on your machines: Windows, Mac OS and Linux alike.
  • The LLM of your choice: various hosted models via api.girard-davila.net, ask about gpu rental on vast.ai to host more models
  • Safe automation: cron, monitoring, alerts, metrics before anything ships.
Deployment proof: Windows computer + deepseek-v4.1-flash (already deployed at a client)

The step-by-step deployment

Create your account at api.girard-davila.net to get your API key (Google sign-in), then connect pi in a few commands:

  1. curl -fsSL https://pi.dev/install.sh | sh
  2. pi install git:github.com/alx/pi-provider-gd
  3. pi -> /login gd # Google sign-in at api.girard-davila.net, key stored automatically
  4. /model gd # pick a model, e.g. gd/deepseek-v4.1-flash
  5. say "test" # the agent replies

For your agents, on the model side

An api.girard-davila.net account: key generated in a few clicks, credits included in beta. One-command connect via pi-provider-gd.

Book the scoping call

Modules

Five training modules, each standalone or combinable into a path. Every module starts from a live demo, not a slideshow.

M1 · AI agent fundamentals

½ day

What actually changes with agents? A shared foundation for the whole team, in decision + team mode.

M2 · Automating your daily work

1 day

From demo to production script: your real tasks, automated at your place, not the trainer's.

M3 · Autonomous pipeline: ralph, beads & sub-agents

2 days

A development loop that works by itself: issue base (beads), autonomous iteration (ralph), sub-agents that check each other.

M4 · Local LLMs & image generation on GPU

1–2 days

llama.cpp on a 12–24 GB card, local OpenAI-compatible API service, Stable Diffusion with control (LoRA, post-processing).

M5 · Agentization audit & roadmap

2–3 days on-site

Map your workflows, score what can be automated now, quantify the gain, produce a roadmap leadership can execute without me.

See the full module details →

Point of contact

Let's talk about your training.

Describe your team and your goal in two lines. You get a free 30-minute scoping call within 48 business hours.

One human on the other end, no qualification bot. If I'm not the right person, I'll tell you and point you somewhere.

✓

Message sent.

I read everything myself. A human replies within 48 business hours.

Goes straight to a human. No bot, no newsletter.