Corporate training · Two programs
Two AI agent programs your company
can actually put to work
Two in-company programs, both already delivered to international groups: a 2-day hands-on technical intensive that takes engineering and data teams from agent fundamentals to production readiness, and an executive program that gives leadership committees the understanding they need to set their AI roadmap.
Two formats
Two programs, both delivered and proven
These aren't syllabi waiting for a first client. Both formats have been delivered to international groups: one to engineering teams, one to an executive committee.
Designing Reliable AI Agents
A 2-day technical intensive: five modules from agent fundamentals to a production-grade design discipline, with hands-on Colab labs throughout. Your team leaves knowing what agents are, when to use them, and when not to, and how to keep them reliable.
See the five modules →AI Strategy for Executive Committees
Delivered to the executive committee of a global insurance group, to its regional CEO and her full leadership team: a no-code program for leaders who need to understand AI, GenAI, and especially AI agents well enough to set their strategic AI roadmap. Frameworks for decisions, use-case selection, and the leadership questions that come with them. Two half-day sessions.
See the executive program →Program 1 · Technical intensive
Designing Reliable AI Agents: what your team learns
Five modules that move from first principles to a production-grade design discipline. No framework worship; your team learns what agents are, when to use them, and how to keep them reliable.
AI Agent Fundamentals & Architecture
LLM foundations, RAG, and the anatomy of an agent, plus the question most courses skip: when NOT to use an agent. Your team leaves able to map a problem to the right planning pattern, or to no agent at all.
Advanced Design Patterns, Frameworks, Reasoning, Memory
Workflow patterns, orchestrator/workers, evaluator-optimizer, multi-agent systems, and a full memory taxonomy. Reasoning techniques: CoT, ReAct, Reflexion, and ToT, implemented across OpenAI Agents SDK, LangGraph, LlamaIndex, and Anthropic SDK.
Evaluation, Observability & Tracing
If you can't measure it, you can't ship it. OpenTelemetry instrumentation, Arize Phoenix tracing, LLM-as-a-judge evaluation, and trajectory metrics, so your team knows what their agents actually did, and how well.
Security, Robustness & Production Readiness
Guardrails, prompt-injection defense-in-depth, least-privilege tool access, retry/fallback/ circuit-breaker patterns, and drift monitoring. The difference between a demo and a system you can put in front of real users.
How to Design an Agentic System
Everything converges into an 8-step design checklist your team applies to their own use cases, anchored by the golden rules: "Start simple. Avoid agents by default. Increase complexity only when justified."
Taught by a practitioner
Packt author of Building AI Agents for Finance, O'Reilly video course instructor, international conference speaker, with 16+ years in financial markets across algorithmic trading, machine learning for finance, and data science & BI team leadership.
Pedagogy
Understand the mechanism, then use the real tools
Hands-on Colab labs, built on a simple conviction: engineers who understand how an agent reasons, remembers, and decides to act can design systems that hold up, and fix them when they don't.
Concepts your team can build with, not just recognize
Reasoning, memory, tool use, architectural styles: each is taught as a mechanism before it is code. What actually steers the model's next step, what memory keeps and what it forgets, where control passes between agents. Then the team implements it in a lab. Nothing is taken on faith; even tool calling is stripped back to the raw exchange, where the model never runs your function, it only asks.
Finance-based labs, transferable patterns
The Colab notebooks are built on finance use cases, the instructor's own domain, so the examples carry real constraints instead of toy data. The design patterns they teach are domain-agnostic: engineers from any industry apply them directly to their own systems.
Multi-framework, no vendor lock-in
OpenAI Agents SDK, LangGraph, LlamaIndex, Anthropic SDK; the same patterns across all of them. Your team learns transferable design skills, not one vendor's API surface.
Production reliability as the through-line
Evaluation, observability, guardrails, and failure handling aren't an appendix; they run through every module, because the goal is agents your organization can actually trust.
Program 2 · For leadership teams
AI Strategy for Executive Committees
Delivered as two half-day sessions to the executive committee of a global insurance group, to its regional CEO and her full leadership team. No code, no notebooks; frameworks for decisions. For leadership teams who need to understand AI, GenAI, and especially AI agents well enough to set their strategic AI roadmap themselves, rather than outsourcing the thinking to vendors.
What AI agents really are, and aren't
A clear, hype-free mental model of LLMs, GenAI, and agentic systems: what they can do today, where they fail, and what separates a demo from a system you can rely on.
Which use cases justify an agent
How to sort your candidate use cases: which ones justify agents, which need simpler automation, and which should wait. The same selection discipline the technical program teaches engineers, translated for decision-makers.
Where AI ROI actually lives
Traditional AI, GenAI, RAG, and agentic systems mapped against the returns actually observed in the field. Why most GenAI initiatives still show no measurable return, and what the ones that work do differently.
The leaders' role, and your roadmap
From the tech-first reflex to business-first reasoning, and the leadership postures a committee needs to balance. Two half-days will not write your AI roadmap, but they leave you with the structure to build it and a prioritized shortlist to build it from.
Why two half-days, and not two days
An executive committee's calendar does not clear for two full days, and it does not need to. Two half-days keep attention high, sit either side of an agenda the committee already has, and leave thinking time in between, which is usually when the real decisions get made. Each half-day is self-contained: the first builds the shared understanding, the second turns it into a scored shortlist and a roadmap. The format extends into an ongoing program when the committee wants to keep going.
Flexibility
Advanced by design, adapted to your audience
The technical intensive is an advanced program, built for teams who will design and ship agent systems. But it adapts: a lighter, less technical variant exists for mixed or business audiences, and between the full intensive and the executive program there's a right depth for every team. Emphasis, pace, and technical depth are tuned on the scoping call; the curriculum flexes to the room, not the other way around.
Process
How it works
From first conversation to post-training follow-up, delivered with the full professional apparatus: formal training convention, detailed pedagogical program, attendance sheets, and completion attestations. French training-organization registration in progress (DREETS Île-de-France).
Scoping call
We map your teams' current level, use cases, and constraints, agree on objectives, and pick the right format: technical intensive, executive program, or a lighter variant.
Calibration
Depth and emphasis are tuned to your audience's level, from executive to advanced engineering. Labs use proven finance-based notebooks; how the patterns transfer to your own domain is covered in discussion, not by rebuilding the labs.
Delivery
On-site or remote, in English or French. The technical intensive runs over two days, combining concise theory with hands-on Colab labs; the executive program runs as two half-day sessions, because a committee's calendar does not clear for two full days.
Follow-up
Hot and cold satisfaction surveys (FR/EN), completion attestations for every participant, and a continuous improvement loop feeding back into the program.
Participant feedback
What participants say
Answers to "what were the strengths of this training?", collected from participants of Designing Reliable AI Agents delivered in-company. Quoted verbatim.
"Good balance between theory and hands-on practice. Focus on production topics. Well structured modules."
"Very complete, draw the boundaries of this domain, allow to see the full aspects of an Agentic project. Both theory and practice."
"Deep knowledge of the trainer, completeness, practice."
"Clear explanations, strong engagement with us."
"Exhaustive, competent trainer, theory + practice."
"Very complete training regarding the Agentic Systems."
FAQ
Common questions
What team size works best?
What are the prerequisites?
Can the content be customized to our context?
Which languages do you deliver in?
What about logistics and pricing?
Next step
Ready to bring it to your teams?
A 30-minute scoping call is enough to tell which format fits your teams, and what the program would look like at your organization. The detailed program, module by module, is shared from there.