Prompts as code, gateways, tracing, evals as CI — the applied layer between a model API and a product people trust. This hub grows out of the old Engatica post that ranked for "llmops" in 2023.
What LLMOps actually means in 2027 — prompts, evals, tracing, gateways and agent operations — and how it grew out of MLOps. The updated edition of Engatica's LLMops post.
A practical programme for teaching a whole company what AI fairness means — updated for the era of LLM agents, from data provenance to incident drills.
The successor to our most-shared trends piece: agents go operational, evals become the moat, small models eat the edge — and five more trends that will decide 2027–28.
What LLM agents are, how they work — tools, memory, planning, guardrails — and a build-vs-buy framework for teams shipping their first production agent.
What 'squinting strategy' means — and how deliberate blur helps leaders read markets, portfolios and AI bets in 2027.
Running the stack yourself? The LLMOps guide maps the layers; some open-source platforms implement several of them in one self-hosted app.
LLMOps is the discipline of operating systems built on models you rent rather than train: versioned prompts, an AI gateway in front of providers, tracing of every call, evaluations run like CI, and operations for the agents built on top. It grew out of MLOps, but the artefacts under your control are prompts, retrieval, tools and guardrails.
An LLM is the model — it predicts text. An agent is the system around a model: tools it can call, memory, planning loops, guardrails and evaluation, arranged so the model can complete tasks rather than just answer questions.
Operators should start with LLMOps in 2027 for the stack; builders should start with the practical guide to LLM agents; teams deploying either should read the fairness guide before launch.