Taking Data Science in industry from zero to production
summary
Describes a workflow that moves projects from prototype to production by validating concrete assumptions rather than polishing features: use verifying income from bank transactions as a worked example, enumerate feasibility value usability viability risks, prioritize the riskiest assumption and test it, aim for about one or two validated assumptions per week when possible, keep delivery scoped only to validated assumptions, use regular peer checkins to avoid rabbit holes, and prefer lightweight prototypes (Figma or simple models) that either fail fast or justify engineering investment.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:02Data science in production struggles
- 2:29Discovery process
- 3:34Business impact outcomes
- 7:30Discovery and delivery cycles
- 10:43Team roles in discovery
- 12:47Managing uncertainty
- 14:57Delivery with validated assumptions
- 16:43Building trust with outcomes
- 19:33When to use discovery
- 20:39Answering audience questions on discovery
related talks
07mThe Art of Problem-Solving: Strategies for Finding Your Way in the Dark
40mBuilding Successful Data Science Projects
42mMasterclass: The right path to ML
28mData Science Project Patterns that Work
09mLeading a team through 0-1 product development
14mGathering Requirements, and Rapidly Building Proof of Concepts
35mApplied Data Science
31mProductionalizing a Data Science Team