ML Feature Engineering

A feature inventory and naming standard, point-in-time correct training joins, skew and parity checks between training and serving, shared transformation code, and the routines that keep a feature store lean and trusted.

Free · Mac (Apple silicon) and Windows · v1.0.5 · Other platforms

Then open the Marketplace in the app, search for ML Feature Engineering, and click Install. It goes into your own files, and works offline from then on.

What it installs.

1

Pillar

1

Area

50

Projects

203

Starter tasks

📊 Data & Analytics

ML Feature Engineering

Designing, computing and sharing the input features that machine learning models depend on, consistently between training and production.

  1. Feature inventory across production models
  2. Feature naming and definition standard
  3. Entity keys and grain map for features
  4. Event timestamps and as-of rules for every input
  5. Point-in-time correct training set joins
  6. Training-serving skew audit on one live model
  7. Freshness requirement for each feature group
  8. Shared feature transformation library
  9. Fixture datasets for testing feature code
  10. Feature code review checklist
  11. Weekly feature pipeline health check
  12. Feature validation rules on every run
  13. Monthly online and offline parity check
  14. Feature backfill runbook and log
  15. Feature definition change and versioning routine
  16. Unused feature sweep
  17. Upstream schema change watch for feature inputs
  18. Feature experiment log
  19. Quarterly feature platform cost review
  20. Feature group owners and handover notes
  21. Categorical encoding without target leakage
  22. Rolling window and recency aggregations
  23. Numeric transforms, scaling and binning
  24. Date, time and calendar features
  25. Missing values as features
  26. Text features from counts to embeddings
  27. Ratio, interaction and domain features
  28. Feature engineering practice on public datasets
  29. Feature store build, buy or skip decision
  30. Feature selection pass on an overgrown model
  31. Ablation study of feature groups
  32. Learned embeddings versus hand-built features trial
  33. Precomputing slow request-time features
  34. Merging duplicate features across teams
  35. Retiring a deprecated feature safely
  36. First model served from the feature store
  37. Historical backfill for a new feature set
  38. Feature platform migration with a parallel run
  39. Pre-launch feature review for a new model
  40. Post-incident review of a broken feature
  41. Solo engineer's first feature pipeline without a store
  42. Researcher turning notebook features into a package
  43. Feature construction record for a published study
  44. Data scientist to platform team feature handoff
  45. First month on a team with a feature store
  46. Cold-start features for a product with little history
  47. On-demand features computed at request time
  48. Shared entity embedding registry
  49. Graph features from relationship data
  50. Automated feature generation with guardrails

Where it starts.

Each project arrives with its purpose, milestones and first tasks written out. Here is how the first three read.

Feature inventory across production models

Most teams that start talking about a feature store cannot say how many features their live models actually consume, where each is computed, or which ones two models compute differently under the same name. A sheet listing every input of every production model, with its source tables, computation code and owner, is the starting point for every later decision in this area.

First tasks

  • List every model currently serving predictions and who owns it
  • Extract the input feature list from each model's saved artefact or serving code
  • Record source tables, computation job and refresh frequency per feature
  • Flag features that appear under different names or code in two models
  • Refresh the feature inventory against the model registry @recurring(yearly)

Feature naming and definition standard

Names like user_score_v2_final tell nobody what was aggregated, over which window, or as of when. Agreeing one naming pattern (entity, measure, aggregation, window) and a short definition card for each feature lets an engineer read a training set without opening the code, and stops two people building the same thing twice.

First tasks

  • Collect twenty current feature names and note what each fails to say
  • Draft a naming pattern of entity, measure, aggregation and window
  • Write definition cards for the ten most used features
  • Get sign-off on the standard from the ML and data engineering leads

Entity keys and grain map for features

Every feature belongs to an entity at a grain: a customer per day, an order at checkout, a device per hour. When those keys are fuzzy, joins fan out silently, training rows multiply and a model learns from duplicated examples. Mapping the entities your models predict about, their join keys and the grain of each feature table removes a whole class of silent errors.

First tasks

  • Write down the entities your models make predictions about
  • Record the join key and grain of every feature table
  • Document how merged or reissued keys are resolved
  • Add a test that fails when a join changes the training row count

Make it yours.

Install ML Feature Engineering, then tell Aurum the date you're aiming for. It checks whether the plan fits the weeks you actually have.

Free · Mac (Apple silicon) and Windows · v1.0.5 · Other platforms

Blueprints are general information to help you organise your own work. They are not medical, financial, legal or other professional advice. See the Terms.