Operationalizing ML - Shreya Sankar

[manisnesan]

• 3 success factors in ML: Velocity | Validate early | Versioning ( Source: Youtube)
• 4 Main tasks in production ML : Data Collection | Experimentation | Evaluation & Deploy | Monitor & Respond ( Source: YouTube)
  • Text book tasks: Data Preparation | Training | Evaluation | Deployment
  • Why it resonates: Hold out Set is not constant. We always look for mispredictions and try to expand our hold out is reflecting that and also incorporate that in our data collection by collecting more training samples either by manually labeling or through other data augmentation.
• Source : 4 main tasks in Production ML Lifecycle, MLOps and Datacentric AI

[manisnesan]

• Gradient Dissent - Podcast
• Twitter Thread Takeaways Outerbounds Podcast