Miscellaneous next steps

[peacfuljoh]

• [x] Set up MongoDB to store raw data examples (NoSQL) (versioned)
• [x] Create API on M3 to implement database exploration + “get” queries
• [x] Create data preparation pipeline (raw features)
  • [x] Set up feature store (versioned) for this on M2
  • [x] Add API call to get features
• [ ] Design ML models to map features to predictions
  • [ ] Classic simple regression model (sklearn, pytorch, custom)
• [ ] Set up ML training
  • [x] Set up model store (with versioning) for this on M2
  • [ ] Implement ML API
• [ ] Set up ML inference
  • [ ] Add API calls (evaluate predictions for specific user/post)
• [ ] Set up ML continuous training
  • [ ] Pull latest data, store it, featurize it, re-train model with it
  • [ ] Generalize train API to enable “regular retraining” job
  • [ ] Automate hyperparameter tuning
• [ ] Implement one-page web dashboard with Flask on M3
  • [ ] View data
  • [ ] View predictions for given user and video
  • [ ] Incorporate Redis for performance
• [ ] Migrate to AWS (containerize?)

[peacfuljoh]

Use Tableau as a BI tool