Platform / SRE

LLMOps & MLOps

What changes operationally when the artifact is a model, and what does not.

Who this is for

Platform, DevOps, and SRE engineers running AI workloads on a rota.

What you should be able to do

Operate a model in production with drift detection, evaluation gates, and a retraining path.

Career ladder

The titles this track maps onto. Levels differ between companies. The useful part is the direction, and what each step adds to the one before it.

DevOps Engineer
MLOps Engineer
Senior MLOps Engineer
Staff LLMOps Engineer
Principal Reliability Architect

Tech stack

What the work is actually done with. Grouped by the job each tool does, so the list reads as a system rather than a pile of names.

Pipelines: MLflowKubeflowGitHub Actions
Registry: Model registryPrompt versioningFeature store
Monitoring: PrometheusOpenTelemetryDrift detection
Release: CanaryShadowRollbackEval gates

The delivery flow

The order the work actually happens in. Each step is where a decision gets made and written down, not a chapter heading.

CI + eval gate
Model registry
Deployment strategy
Serving + autoscale
Telemetry
Drift detection
Retraining trigger
Rollback

Reference repository structure

A starting layout for this track. The directories are the ones that get added late and hurt. Decisions, evals, policy, lineage. Promoted to the top level where they are visible.

mlops/
├── pipelines/
│   ├── train/
│   ├── evaluate/          # gate: build fails on regression
│   └── deploy/
├── registry/              # model + prompt versions, immutable
├── features/              # feature store definitions
├── monitoring/
│   ├── drift/
│   ├── quality/
│   └── cost/
├── release/
│   ├── canary.yaml
│   └── rollback.md        # the runbook, tested
└── runbooks/

Reading path

9 published guides on this track. Every one is a live page on this site. Nothing here is a placeholder.

Reference implementations

Citadel’s open-source repositories for this track. Terraform modules, MCP servers, and reference architectures you can read, fork, and run. Apache/MIT licensed; check each repository for its terms.

Primary sources

The standards, framework documents, and vendor references this track is built against. Go here when you need the authoritative wording rather than a summary. In a security review or an audit, the source is what counts.

Other tracks

Talk through this with an architect