Check: MLOps (mlops) as well.

• Scoping:

  • Define the project.
  • Decide on key metrics.
  • Estimate resources, and timeline.

• Data:

  • Define data.
    • establish a baselines, and address data labeling inconsistencies.
    • obtaining data.
  • Standardize data labeling conventions.
    • Advanced Labeling Technique:
      • Semi-supervised Learning
      • Active Learning
      • Weak Supervision

• Modeling: To get started in modelling there are 3 key input: Code (algorithm/model), Hyperparameter, Data. In general: Optimizing data and hyperparameters can be more effective for production systems. (the code (algorithm/model) is fixed.

  • Select and train the model while optimizing data and hyperparameters.
    • optimizing data such as using
      • data augmentation
      • feature engineering
      • feature selection and dimensionality Reduction
  • Perform error analysis in machine learning to identify areas for improvement.
  • Perform model debugging to ensure model robustness.
  • Improve model robustness by performing model remediation.
  • Use ML Experiment Tracking tools throughout the modeling processes.

• ML Model Deployment:

  • Deploy the system (Perform model resource optimization if needed)
    • On-Edge
    • Online Inference using a service (ML model serving)
    • Batch Inference (ML Batch Inference)
  • Monitor, and maintain it in production.
    • To deal with data distribution changes.

References

1. https://www.coursera.org/learn/introduction-to-machine-learning-in-production/home/week/1
2. https://community.deeplearning.ai/t/mlep-course-1-lecture-notes/54446 (need login)