Six months ago I wouldn’t have called myself an “AI engineer”.

Today I’ve built:

• a self-improving agentic system (Bob + Chad)
• a data-driven arbitrage engine (GhostFrog)
• a full Civil Service AI toolkit
• a Flask portfolio full of working apps
• local LLM pipelines using Ollama
• an entire developer workflow powered by autonomous tools

What changed?

I stopped learning theory.
I started shipping projects that matter to me.

The breakthrough was realising that:

• Python skills grow fastest when you build end-to-end systems
• Agentic patterns (planner → tools → executor) create insane productivity
• Local models (phi, qwen) outperform cloud LLMs for iterative coding
• Every mistake becomes a learning rule for the systsoem

The Big Outcome

I now feel genuinely ready to apply for Python developer roles — including UK Civil Service SEO/G7 roles.

I built my own evidence instead of reading books.

This blog will track the next stage of the journey.