Why I'm learning AI engineering in public
I've deployed an AI agent inside a real company. Now I'm documenting the path from data scientist to AI engineer, one shipped project at a time.
Most of what you read about AI agents stops at the demo. Someone connects a model to a few tools, records a clip of it booking a meeting or summarising an inbox, and calls it done.
The hard part starts after that. Which account does the agent run as, and what is it allowed to touch? What happens when an API times out halfway through a task? How do you know the agent did the right thing on Tuesday at 3am when nobody was watching? And what does it cost per run once real volume arrives?
Where I’m starting from
I’m a Data Science & AI Analyst in London. I studied Computer Systems Engineering at the University of Bath, then an MSc in Management, and for the past two years I’ve worked across three consumer brands building data pipelines, machine learning models and dashboards.
Along the way I designed and deployed an autonomous agent on Google Cloud that automates a recurring compliance reporting workflow, using Vertex AI, the Gemini API and the Gmail API. I’m also one of four internal consultants helping an 800+ person company adopt AI.
That agent taught me more than any course. Most of the lessons had nothing to do with prompts. They were about permissions, reliability and trust.
What I’ll write about
This site is where I’ll document my move into AI engineering, in the open:
- Projects I ship, with the code, a live demo where possible, and the decisions behind them
- Production lessons: service accounts, evaluations, tracing, guardrails and cost control
- Mistakes, because the failures are usually the most useful part
I’ll keep anything about my employer at the level of patterns and lessons, never internal details.
Why in public
Two reasons. Writing forces me to understand something properly before I claim to. And if you’re working on the same problems, whether as an engineer, a hiring manager or a team trying to get agents into production, I’d like to hear from you.
The first project write-up is coming next. If you want to follow along, connect with me on LinkedIn or subscribe via RSS.