01
Choose The Job
Leave with: Agent brief + success test
Turn one useful business problem into a narrow agent mission with clear inputs, outputs, limits, and a human owner.
The First Agent Program
Build Your First Agent
Hands-on online coaching to design, build, secure, and deploy an AI agent that can remember, use real tools, and run beyond your laptop.
See the build path
Live online / Built on your stack
The Build Path
Local To Live
One Agent. The Whole System.
I build with you, in order, so every new capability has a reason to exist. You learn the decisions behind the system while it runs on your real stack.
02
Build The Core
Leave with: Working local agent
Create the instruction layer, model routing, structured state, and control loop. Learn what belongs in prompts and what belongs in code.
03
Give It Tools
Leave with: Reliable tool surface
Connect APIs, databases, files, browsers, commerce systems, or internal services through typed and testable tool contracts.
04
Design Memory
Leave with: Memory policy + storage
Separate conversation context, working state, durable facts, and retrieval. Decide what is remembered, summarized, expired, or deleted.
05
Secure The Work
Leave with: Threat model + guardrails
Add least-privilege access, secret handling, approvals, allowlists, rate limits, audit trails, and a clean way to stop the agent.
06
Ship And Operate
Leave with: Live agent + runbook
Package the agent, deploy it to a VPS, expose only what is needed, and add logs, health checks, backups, monitoring, and recovery.
The two systems most demos skip
Memory With Rules. Access With Limits.
Useful memory and safe action are designed together. The agent should know enough to help—without keeping everything or gaining more access than the job requires.
Memory design
- Context vs. working state vs. durable memory
- Retrieval, summarization, and write policies
- Retention, deletion, and tenant separation
- Memory quality tests and conflict handling
Security design
- Scoped credentials and secret isolation
- Tool allowlists and least-privilege access
- Human approval for consequential actions
- Logs, limits, alerts, and a kill switch
Framework-agnostic by design
Pick The Tool That Fits The Work
You learn how to compare tools by control, reliability, speed, cost, and ownership—not by whichever framework is loudest this month.
01 / Tool layer
Code-first agents
OpenAI Agents SDK, LangGraph, PydanticAI
For explicit control, typed tools, tests, and production ownership.
02 / Tool layer
Automation builders
n8n, Make, visual workflows
For fast integrations, triggers, approvals, and operational handoffs.
03 / Tool layer
Agentic coding
Codex, Claude Code, Hermes-style teams
For planning, implementation, review, and multi-agent build workflows.
04 / Tool layer
Memory and data
Postgres, Redis, vector stores, files
For state, durable facts, retrieval, caching, and auditability.
05 / Tool layer
Production surface
Linux, Docker, Nginx, VPS, monitoring
For secure deployment, uptime, visibility, backups, and recovery.
Beyond localhost
Make It Work When You Close The Laptop.
Deployment is part of the coaching, not a footnote. I move the agent with you to an external environment and make the operating model clear enough for you to maintain.
VPS setupDockerTLSHealth checksLogsBackupsAlertsRecovery
What you leave with
A Live System You Understand.
01
A working first agent tied to a real use case
02
A readable architecture map and tool contract
03
A memory model with retention and deletion rules
04
A security checklist with scoped permissions
05
Evaluation cases for quality and failure modes
06
Deployment files, monitoring, and an operations runbook
Bring a use case or start from zero
Map Your First Agent With Me.
Good to know
Frequent Questions
01How does the quote work?
You won't find prices on this site: every training and every project is quoted for the person, the project and the country. Fill in the form, and within two business days I send you a quote and the next steps by email.
02I'm a complete beginner. Can I follow?
Yes. My trainings start from the job to do, not from theory: you pick a real mission and we build it step by step, live. If you would rather move at your own pace with me alone, the 1:1 coaching exists exactly for that.
03In a group or 1:1?
The group sessions are for building alongside others on the same goal: a first agent in six weeks, or a real app in four sessions. The 1:1 is for your own path — freelancing, a SaaS, agents for clients — in four 60-minute sessions over a month, with messages in between.
04Where and when do sessions happen?
Everything happens online, on Google Meet. I work on Bangkok time (UTC+7), Monday to Friday, 10:00 to 18:00, and we pick the session slots inside that window. You need a laptop and a stable connection, nothing else.
05How do I pay, and how from Algeria?
By bank transfer in euros. A project starts once the 30% deposit has arrived. If you are in Algeria, it works the same way: a transfer in euros to my euro account, so no card and no platform fees in the middle.
06Can I cancel or move a session?
Yes, for free up to 24 hours before the session starts. After that the session counts as done, and a no-show is a lost session. I keep this rule simple so my calendar stays reliable for everyone.
07Which language do you work in?
French and English, both for the sessions and for the written work. Pick whichever you are most comfortable with; I switch between the two without any problem.
08What happens after I send the form?
You get a confirmation email right away. Then I read your answers myself and reply personally, within two business days, with a quote or the next steps. If you would rather talk first, book the free 20-minute call on Google Meet.
The Handoff
Built To Transfer
You Leave Ready To Run It
A working first agent connected to one real job, with clear inputs, outputs, tools, limits, and a success test.
A memory system that separates context, state, durable facts, and retrieval—with explicit rules for what gets stored.
Scoped tools, protected secrets, approval gates, rate limits, logs, and a kill switch you know how to operate.
A deployed service with health checks, monitoring, backups, recovery steps, and a runbook that makes ownership practical.
