CORSO 02 · NOEVA ACADEMY
From Developer to FullAI Developer
Beyond using AI as an assistant: design, integrate and orchestrate agents inside your software.
Developers and technical teams (back-end, front-end, full-stack), tech leads and DevOps.
Programming in at least one language (Python or JavaScript/TypeScript preferred), Git and command line, basic REST API concepts.
Beyond using AI as an assistant: design, integrate and orchestrate agents inside your software.
Hands-on on your own repo
Each participant works on their own repository with a real dev environment: not slides, code.
From API to custom MCP server
Integrate a model via API, build Agent Skills and MCP servers to connect AI to external systems, end-to-end.
Quality, evals and security
Testing non-deterministic components, evaluations, prompt injection, secrets and cost control.
Assessed capstone project
Close with a complete FullAI application, with custom tools/MCP, evals and hardening, assessed on functionality and security.
- Adopt the mindset of AI-agent-assisted development (“FullAI”).
- Master prompt engineering applied to code and technical tasks.
- Integrate AI models into your apps via API (tool use, streaming, multi-turn conversations).
- Use a coding agent productively and safely across the development cycle.
- Create reusable Agent Skills and MCP servers to connect AI to external systems.
- Orchestrate sub-agents and manage context in complex flows.
- Apply testing, evals, security and CI/CD to AI development.
By the end, participants can design and build an application that integrates an AI model, extend its capabilities with custom tools and MCP servers, automate parts of their development work with agents, and assess the quality, cost and security of the solutions they build.
9 modules · 37 lessons
Module 01 FullAI mindset: from manual code to agents 2 hours · 4 lessons
Objective Redefine the developer’s role in the age of AI agents.
- 1.1 What changes
From writing every line to directing and validating an agent’s work; the developer’s new responsibilities.
- 1.2 Taxonomy
Assistant, copilot, autonomous agent, multi-agent system: differences and use cases.
- 1.3 The Explore-Plan-Code-Verify flow
The working cycle with a coding agent and why it works.
- 1.4 Ethics and technical limits
Reliability, cost, security and ownership of generated code.
Module 02 Prompt engineering for developers 3 hours · 4 lessons
Objective Write effective instructions to generate, review and fix code.
- 2.1 Technical prompts
Specify language, constraints, style, dependencies and acceptance criteria.
- 2.2 System prompts and roles
Set the model’s behaviour; persistent vs session instructions.
- 2.3 Advanced techniques
Few-shot examples, step-by-step reasoning, structured tags (e.g. XML), constrained output.
- 2.4 Anti-patterns
Ambiguous prompts, too much or too little context, over-engineering the request.
Module 03 Integrating AI with APIs 6 hours · 5 lessons
Objective Call an AI model from code and handle its responses in a real application.
- 3.1 API fundamentals
Authentication, message structure, parameters (temperature, max tokens), error and rate-limit handling.
- 3.2 System prompts and multi-turn conversations
Maintain state and context in an application dialogue.
- 3.3 Streaming
Receive responses in real time and improve the user experience.
- 3.4 Tool use / function calling
Let the model call your own code functions; tool schema and result handling.
- 3.5 Cost and observability
Estimate and contain cost; logging and monitoring of calls.
Module 04 Agentic coding with a terminal agent 4 hours · 4 lessons
Objective Use a coding agent to explore, change and maintain a real codebase.
- 4.1 Setup and project context
Configure the agent; project instruction files and repository conventions.
- 4.2 The Explore-Plan-Code-Verify cycle
Work on features and bugs by guiding the agent step by step.
- 4.3 Refactoring and tests
Generate and maintain tests; safe refactoring on existing code.
- 4.4 Control and review
Read diffs, commit deliberately, keep human control over changes.
Module 05 Reusable Agent Skills 3 hours · 4 lessons
Objective Create reusable skills that teach the AI recurring tasks.
- 5.1 What a Skill is
The idea of reusable instructions (SKILL.md) applied automatically to the right task.
- 5.2 Anatomy of a Skill
Format, the description that governs activation, linked resources and scripts.
- 5.3 Dynamic context injection
How the skill is discovered, selected and invoked.
- 5.4 Sharing
Distribute skills across the team via repositories and packages.
Module 06 Model Context Protocol (MCP) 5 hours · 4 lessons
Objective Connect AI to external data and services by building an MCP server.
- 6.1 Why MCP
The standard to connect models to tools and data; client-server architecture.
- 6.2 The three primitives
Tools, resources and prompts: what they are and when to use them.
- 6.3 Building a server
Implement an MCP server from scratch and expose tools and resources.
- 6.4 Transports and connection
Transport mechanisms; connect the server to an AI client.
Module 07 Advanced MCP, sub-agents and orchestration 3 hours · 4 lessons
Objective Handle complex scenarios with advanced patterns and delegation between agents.
- 7.1 Advanced MCP patterns
Sampling, notifications, file-system access and production patterns.
- 7.2 Sub-agents
Delegate tasks to isolated sub-agents to reduce noise and manage context.
- 7.3 Orchestration
Coordinate multiple agents/tools in a flow; passing state and results.
- 7.4 Context management
Strategies for limited context windows and long-running tasks.
Module 08 Quality, evaluations and security 3 hours · 4 lessons
Objective Ensure reliability, security and cost control in AI solutions.
- 8.1 Testing AI software
Test non-deterministic components; testing strategies and test data.
- 8.2 Evaluations (evals)
Define metrics and datasets to measure output quality over time.
- 8.3 Security
Prompt injection, secret management, least-privilege for tools and MCP.
- 8.4 Cost and performance
Model choice, caching, token and latency optimisation.
Module 09 Capstone project 8 hours · 4 lessons
Objective Bring all skills together in a complete, assessable project.
- 9.1 Project definition
Choose a real case: an AI application with custom tools/MCP.
- 9.2 Guided development
End-to-end build with trainer support: API, agent, skill, MCP.
- 9.3 Evals and hardening
Add evaluations, tests and security measures.
- 9.4 Demo and final code review
Presentation, peer review and company adoption plan.
Hands-on with a development environment. Each participant works on their own repository.
NOEVA Academy certificate upon completing the capstone project.
Bring AI into your team
Tell us the roles, goals and size of your team: together we build the right edition, in the classroom or remote.