Data labelling to scale your ML and AI pipelines
summary
Covers building an enterprise-grade data labeling/annotation pipeline to support ML/AI systems, focusing on how to make annotation a disciplined, scalable part of the development lifecycle from PoC to production. Describes considerations, best practices and design pointers for creating iterative, high‑quality data annotation practices that avoid becoming a production bottleneck.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:47Obsession + Craft
- 2:49Software 2.0
- 3:21Algorithm Training is Algorithm Design
- 4:41Data Annotation Needs Are Substantial
- 9:40The Case For Enterprise Annotation
- 17:01HELPING CHICAGO CUBS WIN WORLD SERIES
- 18:40Collaborative Framework
- 22:11Good Annotation Design: Context Matters
- 22:51Good Annotation Design: UI Matters
- 23:21Good Annotation Design: Domain Specific
- 24:02Good Annotation Design: Allow Open Feedback
- 26:23In-House Technology Development
- 33:30Summary - Mindful Data Annotation
related talks
50mSpeeding up data labeling with neural nets: pipeline, metrics and life hacks
37mBest Practices for Building AI Datasets
38mAutomatic DB annotation with ML tools and its problems
17mYour AI Deserves Better Data
31mReliable Data for Large ML Models: Principles and Practices
14mScaling ML platforms to industrialize AI/ML development
04mLabel Studio
34m6 Tips to Avoid Pitfalls in Data Annotation