
ML Project Failure Modes and Bright Spots

Building Effective and Thriving Machine Learning Teams - David Tan & Dave Colls
Topic 4 of 9
Published Jan 27, 2025 · Clip 15:01–23:32 (8:31)
The guests discuss common ML project failure modes: 'POC hell' where teams build prototype after prototype that never reach production, and shipping a first version held together by 'duct tape and sticky gum' that is hard to test or change. David Colls adds diffusion of responsibility across teams, illustrated by post-ChatGPT LLM curiosity with misaligned roadmaps and ungoverned data. They contrast these with success tactics: user testing before expensive MVPs, cross-functional teams via the inverse Conway maneuver, and the confidence of CI/CD and production monitoring.