Internship: Agentic AI

← All positions Internship: Agentic AI & Intelligent Systems Engineering Agentic AI / Casablanca, Morocco / Final-Year Internship (PFE)

Location:
Casablanca, Morocco Company: Cognitive Machines Technologies (CMX) Period: February to June 2027 Duration: 4 months | Final-Year Internship (PFE) AI systems are evolving from generating answers to planning, using tools, making decisions, and executing complex tasks. But building an agent that works in a demonstration is very different from engineering a system that can be trusted in a real enterprise environment. At CMX, we want to explore that difference. We're looking for students who enjoy programming, artificial intelligence, algorithms, and understanding how complex systems behave. Research Areas
• LLMs, SLMs, reasoning models, and cognitive architectures.
• Autonomous agents, multi-agent systems, and orchestration.
• Agent memory, planning, tool calling, and MCP.
• RAG, semantic search, and knowledge graphs.
• State machines, deterministic workflows, and human-in-the-loop systems.
• AI-assisted software engineering and legacy modernization.
• Agent evaluation, reliability, observability, and performance optimization. Your Research Mission Using our CMX Cognitive Framework, you'll study existing agent architectures, identify their limitations, design an approach, implement a prototype, and evaluate its reliability through reproducible experiments. We want you to go beyond connecting APIs and calling LLMs. We want you to understand why your system works, when it fails, and how to improve it. Three Questions Before You Apply
• The Autonomous Agent Paradox An AI agent successfully completes 95% of its assigned tasks. However, the remaining 5% can cause serious business damage. Would you deploy it? Under what conditions?
• Reasoning Versus Execution Two AI agents solve the same problem correctly. Agent A uses 20 reasoning steps. Agent B uses only 3. Which one is better? How would you design an experiment to support your answer?
• The Infinite Loop An autonomous agent repeatedly calls the same tools without making progress, but continues reporting that it's working. How would you detect the situation, stop the loop, and allow the system to recover safely?

How to Apply
Please send the following to contacts@cmxarchi.com:
• Your CV.
• A Word document (.docx), maximum two pages, containing your answers to all three questions.
• Your GitHub profile, portfolio, or relevant projects, if available. You may use AI tools and external resources, but you must explain your own reasoning and disclose the tools or sources used. Important: Applications without answers to all three questions in the requested Word format will not proceed to the next selection stage. Shortlisted candidates will be invited to the CMX Talent Acquisition Day in early January 2027 for an in-person assessment and interview. We're not looking for someone who knows every AI framework. We're looking for someone who enjoys understanding complex problems and turning ideas into reliable engineering systems. Interested? Send your CV and a short note to contacts@cmxarchi.com