AI engineering
Practical LLM integration, agents, pipelines, and engineering patterns.
5 materials in collection
Open-weight models and the future of AI infrastructure
Why open-weight models will capture a large part of the enterprise AI market, and which infrastructure layer businesses will actually pay for.
Why AI agents need sandboxes
How isolated environments let agents execute generated code safely, and when open-ended CLI access is more useful than a rigid set of MCP tools.
The most useful post you will read about custom MCP servers
Why wrapping every REST endpoint one-to-one as an MCP tool creates fragile agents, and why tools should represent business actions instead.
AI engineering is the most candidate-friendly market in tech
Why AI engineering currently favors candidates, how its two main profiles differ, and where the market may converge.
BERT on steroids, or a new chapter in AI engineering
Why JEV makes classification with prompt-defined categories fast, cheap, and surprisingly flexible compared with BERT and generative models.