Tooling without training
Most engineers learned agents on the job. They know the interface, not the operating model behind the harness.
AI engineering training
Stop tokenmaxxing. Put engineers back in control.
Two hands-on formats for software teams that need better context-window discipline, fewer blind agent loops, lower token waste, and stronger engineering review.
The problem
Your engineers think they control AI. Your token bill says otherwise.
Context without discipline
Oversized prompts, vague tasks, and polluted context windows make the model expensive before they make it useful.
Usage without accountability
If nobody can explain when to stop the assistant, the cost curve becomes the first visible signal of lost control.
Token consumption is the worst metric software engineering could optimize for.
Who decides
AI governance is now an executive responsibility.
Governance goals
Tech C-level
Make AI a disciplined engineering tool again.
- Protect quality, tests, code review, architecture, CI/CD, and deployment workflows
- Move engineers from AI validators back to owners of the SDLC
- Rebuild team cohesion, knowledge sharing, engineering craft, and talent retention
Other C-level
Govern AI adoption without paying the hidden cost.
- Turn AI productivity into governed leverage, not uncontrolled token spend
- Reduce strategic, security, financial, and organizational risk
- Avoid technical debt from poorly piloted AI transformation programs
Formats
Choose between deep practice and fast team alignment.
Custom intra-company workshop
Controlled Agentic Engineering
A hands-on training path for cohorts of 10 to 15 software engineers, adapted to your use cases, stack, codebase reality, and delivery rituals.
- Master context windows, context packs, and task boundaries
- Understand harnesses, agents, skills, prompts, and thinking-token tradeoffs
- Practice planning, validation, tests, review, and controlled delegation
Internal masterclass / meetup
Tokenmaxxing Rehab
A 3-hour on-site session to create a shared wake-up call, align engineers on good AI usage, and introduce controlled context discipline.
- Expose tokenmaxxing, context dumping, and fake AI velocity
- Give engineers a shared language for better agent usage
- Create momentum before a deeper workshop or internal rollout
Decision point
Not sure which format fits? Start with a 30-minute intake.
We map your team size, current AI usage, and delivery pain to the right entry point.
Signals
Built from real software delivery work, not generic prompt training.
Method
Bring software engineering fundamentals back into the AI loop.
Diagnose current AI usage
Identify where assistants help, where they hide rework, and where teams have no shared operating rules.
Control the context and the harness
Teach how context windows, prompts, skills, tools, agents, and thinking tokens actually shape output quality.
Practice deliberate AI-assisted engineering
Engineers work on realistic flows: debug, refactor, feature work, tests, review, validation, and stopping criteria.
Codify team protocols
The team leaves with habits managers can inspect and engineers can apply after the session.
Your instructors
Practitioners first. Trainers second.
Manuel Odendahl
AI engineering practitioner, remote from the US for demos.
25-year software veteran specializing in embedded and systems programming. Manuel spoke at AI Engineer on LLMs for the working programmer and has used LLMs intensively in day-to-day engineering since 2022.
- Remote facilitator
- 25 years in tech
- Architect · senior software engineer
- AI Engineer speaker in San Francisco and New York
- Coding intensively with AI since 2022
Pierre Vannier
On-site facilitator, room lead, and engineering workflow coach.
Software engineer, entrepreneur and CEO of Flint, Pierre has spent 25 years helping tech teams improve engineering practices. He co-founded the Montpellier GenAI meetup, a community of more than 1,300 members.
- On-site facilitator
- 25 years in tech
- Senior developer · CEO at Flint
- Former pedagogical director at Epitech
- Podcast: IA pas que la Data
- AI ambassador · Plan Osez l'IA
Next step
Start with a 30-minute intake.
We qualify team size, current AI tools, maturity, delivery pain, decision makers, and whether the right entry point is a workshop or a masterclass.
Bring five facts: team size, AI tools in use, main delivery pain, target date, and who sponsors the initiative.
Request the intake