Agile Strategies for High-Impact Data Science & Machine Learning Projects
summary
Argues that treating data work as experiments with timeboxed experiments and synchronous demos shortens cycles, separates research from implementation, and prevents the “data collection rabbit hole.” Recommends Kanban over Scrum for DS to visualize flow, limit WIP and prioritize discovery; shows a concrete pattern—8–10am peak, 2–4pm bump—from a coffee‑shop case, warns notebooks often fail in production, and prescribes hypothesis tests, small deployable models, and just‑in‑time engineering.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 1:11Agile meets data science
- 2:35What data science delivers
- 4:44Coffee shop workflow example
- 8:54Kanban for data science
- 10:17Defining data science goals
- 12:00Data science backlog structure
- 14:26Assumptions in data science
- 15:57Pitfalls in data science projects
- 17:45Fixing data delays
- 19:53Avoiding rabbit holes
- 21:25Domain collaboration
- 22:27Synchronous demos
- 24:54Storytelling for teams
- 26:17Kanban recap
- 28:48Sprint goals
- 30:11From rigid to flexible planning
- 31:56Future of agile data science
- 33:19Q&A
related talks
16mRedefining Agile for Data Science: A Team First Approach
28mHow to manage data-related projects and not fail (too often)?
40mData-Driven Agility
32mWhy and How to Craft Bespoke Data Science
29mData Science Best Practices : From Proof of Concepts to Production
54mThe secret sauce of data science management
01h 04mIs Agile Still the Answer? Lessons from Real World Agility
28mData Science Project Patterns that Work