Edge of Context

Practical AI engineering by Slava Dubrov. I write about the parts of AI systems that have to survive production: agent runtimes, memory, security, retrieval, evaluation, LLM infrastructure, and the developer tooling around them.

Start here

• Agent architecture: reasoning loops, memory, tool use, security, and long-running runtimes.
• Retrieval and evaluation: RAG evaluation, search ranking, and context engineering.
• LLM infrastructure: fine-tuning, vLLM structured outputs, LoRAX serving, and LLM engineering concepts.
• Developer tooling: uv on macOS, pyproject.toml, and local LLMs on macOS.

Recent writing

The full archive is on the blog index. For author background, conference talks, and topic coverage, see About Me.