Agile software development delivers working software early, keeps customers involved, and accepts change as the team learns. It contrasts with planning extensive requirements upfront and building components that cannot yet be used. Early delivery gives customers something useful while there is still time to adjust the work. section 1.3
Choose a working structure that suits the team and its customers. Scrum organizes work into sprints; Kanban uses continuous delivery. Neither is the only valid day-to-day practice. If an organization mandates an approach such as SAFe, account for that constraint. In every case, agree on the process, tools, standards and working practices, and documentation. section 4.3.1
For a sprint-based machine learning project, make the work visible and reviewable:
- Keep a backlog of tasks and specify how each task will be approached. section 1.4
- End each sprint with a team checklist covering completed tasks and required documentation. section 1.4
- Expect tasks to carry into later sprints and add work uncovered during the project. section 4.3.1
Keep the resulting documentation as a record the team can share and reuse. It also supports maintenance and governance once the system is in production. DevOps extends the delivery perspective by connecting developers with the people who operate the software, reducing the mismatch between development assumptions and production reality. section 1.4 section 1.3
