Course overview

AI has moved from autocomplete to systems that plan, write and change code across a whole repository. That shifts where engineering effort goes: less typing, far more work on specification, context and verification. Over two days we work through the full lifecycle - what AI genuinely accelerates, where it quietly introduces risk, and which parts of conventional engineering discipline become more important rather than less.

You will prepare a repository so coding agents can navigate it, write specifications that hold up as a source of truth, delegate bounded tasks to agents, and build the verification loops that make generated changes reviewable. We also cover the security surface that comes with automated code changes, and how all of this lands in an existing git, pull request and CI/CD workflow.

Agenda

The AI-native development lifecycle - Where AI changes requirements, design, implementation, testing, review and maintenance - and where conventional engineering discipline remains essential.
Vibe coding: strengths and limits - Use rapid AI-assisted prototyping intentionally while recognizing the failure modes of poorly specified, weakly verified development.
Repository context engineering - Expose architecture, conventions, dependencies, constraints and project instructions so coding agents can work effectively in existing codebases.
Specification-driven development - Turn goals into explicit specifications, acceptance criteria and executable tasks that guide implementation and provide a stable source of truth.
Agentic development patterns - Delegate bounded engineering tasks, decompose work, coordinate parallel activities and manage hand-offs between automated and human contributors.
Verification loops - Combine tests, static analysis, reviews, evaluation criteria and iterative correction to verify generated changes instead of trusting them by default.
Secure AI-assisted coding - Manage secrets, permissions, external tools, untrusted content, dependency risk and the blast radius of automated code changes.
Team and delivery integration - Integrate AI-assisted development with version control, pull requests, CI/CD, documentation and shared engineering standards.

You do not need to install anything, bring your own API keys, or get tooling approved by your employer. We provide the lab environment, the coding agents and the model access, so nobody spends the first morning fighting a corporate proxy.

Who this course is for

This course is for developers, architects and tech leads who ship production software and want to use AI as a controlled part of their delivery process rather than an unpredictable one.

Outcome

Participants can structure an AI-assisted development workflow that produces reviewable, testable changes and scales from individual coding assistance to controlled agentic engineering.

What's included

Digital documentation.
Access to a virtual training center.
Certificate of completion from croit.

Réserver une place pour le 14 déc.

Facturation