LLMOpsThe LLMOps lifecycle · MLflow 3 · FastAPI · Docker · AWS/Azure
Dashboard
0%
1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
41 small wins to finish your pathNext lesson →

The roles on an LLMOps team

An LLMOps project is a team effort, and each lifecycle stage has an owner: the product owner and business analyst scope it, data people prepare the data and build, and ML and ops engineers evaluate, deploy and run it.

Last updated: 15 Oct, 2026 · LLMOps

The LLMOps project lifecycle laid out the eight stages. This lesson puts people against them. The LLMOps Project Lifecycle video names roles as it walks the stages, and the pattern is that different stages are owned by different people rather than one person carrying the whole pipeline.

Who owns which stage

  • Product owner. Owns planning and requirements, and the call on what the application is for and what success means.
  • Business analyst. Works alongside the product owner on requirements gathering, turning a business need into something the team can build against.
  • Data scientist, data engineer and data analyst. Own data collection and preparation, and shape the build, whether that is a retrieval pipeline or a fine-tuning set, against the use case.
  • ML engineer. Owns the model work: fine-tuning or the retrieval build, prompt iteration and the evaluation metrics that score the result.
  • Ops engineer. Owns deployment, the API and scaling, monitoring and the cost of API calls, on a cloud or in the company's own infrastructure.
Three horizontal lanes, one per group of owners: the product owner and business analyst, then the data and ML engineers, then the ops engineer with the ML engineer. The eight lifecycle stages are placed in the lane that owns them and joined left to right, so the handoffs show as steps between lanes: planning sits with the product owner; data, model selection, the build, evaluation and prompts sit with the data and ML engineers; and deployment and monitoring sit with ops.

Roles against the lifecycle

StageWho leads
Planning and requirementsProduct owner, business analyst
Data collection and preparationData scientist, data engineer, data analyst
Model selection and buildML engineer, data scientist
Prompt engineering and evaluationML engineer
Deployment and scalingOps engineer
Monitoring, maintenance, versioningOps engineer, ML engineer

Where the roles matter

  • Staffing a project, where the stages tell you which skills you need and which stage is unowned today.
  • Handing work across a boundary, where knowing that data people pass a build to an ML engineer, who passes a deployed app to ops, sets the seams to watch.
  • A small team wearing several hats, where one person may cover many roles but still has to do each stage's job, not skip it.
Watch out. On a small team one person often plays several of these roles at once. That is fine as long as every stage still has an owner. The failure is a stage with no owner at all, which is how evaluation or versioning quietly gets skipped.
Try it yourself
  • For a project you know, write each lifecycle stage and the name of the person who owns it today.
  • Mark any stage with no clear owner and decide who would take it.
  • Watch the lifecycle video and list every role it names as it moves through the stages.
PreviousThe LLMOps project lifecycle

This is what real progress feels like.