Blog

Notes on reliable AI.

LLM reliability, hallucination mitigation, agentic system design, and what actually works when AI meets regulated industries.

September 3, 2026·2 min read

Engineer or Architect? What 'Architecting AI Systems' Actually Means

The difference between AI engineering and AI architecture, explained through a real project — an orchestrator-free multi-agent system called Mycellium.

AI Solutions ArchitectureMulti-Agent SystemsCareerRead post

September 3, 2026·3 min read

Casablanca to Munich: An AI Student's Path

How a Computer Science M.Sc student from Casablanca, Morocco ended up practicing AI Solutions Architecture in Munich, Germany — and what the move actually changed.

PersonalMoroccoMunichAI Solutions ArchitectureRead post

September 3, 2026·1 min read

The State of AI in Morocco

A look at Morocco's growing AI and tech education scene — UM6P, ENSIAS, and where it connects to the wider European AI conversation from Munich.

MoroccoAI EcosystemRead post

August 3, 2026·1 min read

Why hallucination mitigation matters more than model size

Bigger models hallucinate differently, not less. The real gap is between systems that verify their output and systems that don't.

LLM ReliabilityHallucination MitigationModel SelectionRead post

July 29, 2026·2 min read

Orchestration vs. emergence: when multi-agent systems fail

Both orchestrated and emergent multi-agent architectures fail in production. Here's when to use each and what matters for making them trustworthy.

Agentic SystemsMulti-Agent ArchitectureSystem DesignRead post

July 22, 2026·1 min read

RAG systems still hallucinate. Here's what stops it.

Most people think better retrieval fixes hallucination. It doesn't. The fix is bounded generation, confidence thresholds, and explicit fallback chains.

RAGHallucination MitigationLLM ReliabilityRead post

July 15, 2026·3 min read

Prompt Risk Density: measuring hallucination risk early

Most hallucination research fixes the model. My bachelor's thesis, Echo, measures how much risk is already in the prompt, before generation starts.

LLM ReliabilityHallucination MitigationRead post

July 15, 2026·3 min read

Consensus without an orchestrator

Multi-agent systems usually put a coordinator in charge, and that coordinator becomes the bottleneck. Mycellium works differently: an auction, a blackboard, a consensus rule.

Agentic System DesignMulti-Agent SystemsRead post

Prefer a feed? Subscribe via RSS.