Is AI the right tool? Start with the problem, not the model
Before choosing a model, define the problem, its cost and how you will know it worked. Many tasks are best solved with ordinary code.
Blog
Fourteen components, one post each
Short, practical summaries of every chapter of the book, with the key decisions and takeaways.
Design the workflow before you write a prompt
A ten-step method for designing AI workflows, from splitting the work into steps to the hooks that keep agents in check.
Trusted data in, before any model sees it
Connectors, extraction, validation and quality scoring that turn messy business data into input an AI application can trust.
Know that it works, and keep knowing
Golden datasets, graders and evaluation gates that turn "it looked fine in the demo" into measurable quality.
Interfaces people trust and can correct
Chat is not the only answer. Choosing the right interaction pattern and designing for trust, streaming and correction.
The right store for every kind of data
Relational, object, vector, graph and telemetry stores, and why "Postgres plus object storage" is the right place to start.
Containerise once, deploy anywhere
Local servers, PaaS, cloud VMs, Kubernetes and on-premise, and the habits that keep every option open.
Where the model runs, and why
Frontier APIs, cloud AI services, local models and dedicated GPUs, combined behind one model gateway.
Safe, compliant and in control
Prompt injection, excessive agency, data leakage, layered guardrails, GDPR and the EU AI Act.
See every request, error and resource
OpenTelemetry, structured logs, resilient error handling and Grafana dashboards for AI applications.
See what the agent did, and why
Tracing LLM calls, tools and human decisions, catching silent failures, and closing the loop from feedback to fix.
Quality per euro, by design
Selecting models on your own data, routing to the cheapest capable model, calculating cost per task and optimising without losing quality.
When prompting is not enough
When fine-tuning is worth it, how to build the dataset, how to train locally or remotely, and how to prove it helped.
Every change versioned, tested and reversible
Version control, CI with evaluation gates, prompt and model versioning, progressive releases and instant rollback.