How I think about AI engineering
AI is a systems problem.
I don’t see an AI application as a model wrapped in an interface.
The model is one component inside a larger system — surrounded by context, retrieval, tools, orchestration, memory, business rules, evaluation, observability, and failure handling.
The interesting engineering work is designing how those pieces work together.
Model
- Context
- Knowledge
- Tools
- Memory
- Orchestration
- Evaluation
- Product
The shift
From integrating AI to engineering AI systems.
My early AI work was about integrating intelligent capabilities into existing products.
Over time, my focus shifted toward understanding and engineering the systems around those capabilities — retrieval, agent workflows, context, memory, tools, evaluation, and production reliability.
That shift changed how I approach AI projects.
I no longer start with
- “Where can we add an LLM?”
I start with
- “What system are we trying to build, and where does intelligence actually belong?”