Download the certification brochure (Spanish): Click here

Exam Details

• Exam Name: Agentic AI Certified Expert (AICE)
• Exam Format: Multiple choice, online with remote supervision (proctoring)
• Exam Delivery: Safe Exam Browser
• Number of Questions: 50 questions: 70% scenarios / 30% conceptual
• Exam Duration: 90 minutes
• Passing Score: 70% (35 out of 50 correct answers)
• Available Languages: Spanish / English
• Attempts: Maximum 2 attempts (minimum 14 days between attempts)
• Validity: 3 years · 15 CPD credits per year

Main Subjects

• Generative AI fundamentals (6%): Context engineering; Limits and human-in-the-loop; Value metrics.
• Architecture and model selection (7%): Tokenizer, embeddings and transformer; Inference and context window; Operational evaluation.
• Individual agents (7%): Operational interface; Tool ecosystem; Selection criteria.
• Agentic system design (18%): OODA loop and FSM; Router, Planner-Executor and Critic Loop; Guardrails and observability.
• Shared memory architecture (14%): Append-only journal; MCP, LangMem, Redis and PostgreSQL; A2A and CrewAI.
• Agentic FinOps (5%): Anatomy of cost; Routing, caching and model mix.
• Secure agentic microservices architecture (7%): Zero Trust and mTLS; Least privilege; Policy engine.
• Ethics, GDPR and AI accountability (6%): GDPR principles; DPIA; Accountability.
• Agentic RAG (15%): Chunking and embeddings; Hybrid retrieval and reranking; CRAG and Self-RAG.
• Offensive security and agentic red teaming (15%): Prompt injection; OWASP LLM Top 10; Defense in depth and ASR.

Who is this certification for?

• Platform architects and AI/LLMOps engineers: Who design multi-agent systems.
• FinOps and offensive security leads: Who control costs and risks.
• Innovation leaders and AI governance consultants: Who bring agentic AI into production.

Prerequisites

Mandatory: Agentic AI Certified Associate (AICA). Experience with LLMs, Python and software architecture is recommended.