Data Prep Ops: The Missing Piece of the MLOps Puzzle
summary
Claim: poor AI ROI comes from failing to treat dataset work as engineering. Company Alectio builds a pipeline that prioritizes data selection (useful vs useless vs harmful), merges active learning with training, and routes hard cases via human-in-the-loop workflows. Presenter cites that 90% of firms invest in AI but under 40% see returns, explains synthetic data needs real training data, and shows that trimming harmful/redundant records speeds training and improves model performance.
AI-written summary · about
chapters (20)
- 0:00Why data prep matters
- 1:55Deep learning history
- 4:03MLOps ROI crisis
- 5:27Three pillars of ML success
- 7:51ML lifecycle complexity
- 8:11What is data prep
- 10:40Data prep as flour analogy
- 12:24Data acquisition challenges
- 13:49Synthetic data limits
- 15:17Data scavenging options
- 16:19Data labeling complexity
- 19:54Data transformation basics
- 20:16Data triaging explained
- 21:42Data selection value
- 22:46Data prep ops market gap
- 23:50The boiling frog problem
- 25:55Active learning integration
- 27:19Human-in-the-loop process
- 28:45Concurrent data prep strategy
- 29:05Data as dynamic asset
related talks
26mAre you ready for MLOps? 🫵
32mMLOps, governance and trust
43mNavigating the MLOps Journey: Key Considerations for Successful Implementations
24mFrom DevOps to MLOps: Enabling Artificial Intelligence
13mMLOps at Scale
20mBringing CI/CD Practices to Machine Learning with MLOps
26mThe Hidden Foundation: How Effective Data Management Determines AI System Success
41mSponsored Session: MLOps: Why DevOps Solutions Fall Short in the Machine Learning World