Working Notes on AI Systems

Systems architecture and AI engineering notes on AI agents, LLMs, machine learning, guardrails, and the EU AI Act — from production work.

---

Welcome

These notes capture architecture decisions, engineering patterns, and the trade-offs that come with production AI systems.

This is a personal project that documents my research, experiments and practical experience in Artificial Intelligence.

Its purpose is to organise my research, validate ideas through experimentation and document the engineering approaches I investigate, implement and evaluate in my day-to-day work.

If you spot an error, disagree with an approach or have a better idea, I would genuinely appreciate your feedback.

---

Learning Paths

Curated reading orders — start here.

• LLM in Production
• Agent Architecture
• Clean Architecture for AI
• EU AI Act Compliance

---

Recent publications

• Building a Reliable ML Model — 7 November 2026
• How will people become seniors if nobody gives them a junior role? — 3 November 2026
• What Happens When Agents Evolve, Inherit and Replicate? — 13 October 2026
• How to Use AI Tokens Wisely: A Practical Guide to Better Results and Lower Costs — 23 September 2026
• Building Trustworthy AI Agents: Design Patterns, SHAP and the EU AI Act — 22 September 2026
• Architecture Wars Are Not Just Human Anymore — 19 August 2026
• The Three Measures of Central Tendency — 12 August 2026

Topics

• Large Language Models — cost predictability, LLM FinOps, production deployment
• AI Agents — architecture, protocols, design principles, self-improvement
• Systems Architecture — SOLID principles for AI systems
• Machine Learning — algorithms, clustering, KNN, SHAP, fairness
• Guardrails — prompt injection, safety layers, secure RAG
• AI Act — EU AI Act, risk management, HITL, logging, auditability