Naming conventions
| Element | Convention | Examples |
|---|---|---|
| Entity | singular noun | customer, card, store_sku |
| Feature view | <entity>_<topic> | customer_activity, card_velocity, store_sku_sales |
| Feature | <measure>_<window> or <measure>_<qualifier> | logins_30d, amount_sum_24h, days_since_last_login, plan |
| Window suffix | _<n><unit> with m/h/d/w, or _lifetime | _15m, _24h, _7d, _4w |
| Counts | n_ prefix or _count suffix, consistently | n_txn_1h |
| Distinct counts | nuniq_<thing>_<window> | nuniq_merchant_7d |
| Ratios | <a>_per_<b> or <a>_vs_<b> | spend_per_login_30d, amount_vs_mean_24h |
| Flags | is_ / has_ | has_failed_payment_30d |
| Units | explicit when ambiguous | amount_usd, latency_ms |
| Feature service | <model>_v<n> | churn_model_v1 |
Documentation per feature
- What it measures and the as-of rule ("in the 30 days before event_ts").
- Source table and transformation (link to code).
- Null semantics (no events -> 0 for counts, null for means).
- Owner and creation date; known caveats (changed definition on 2026-03-01).
Ownership
- One owning team per feature view, on call for its freshness SLA.
- Consumers register through feature services, so owners know who breaks when a feature changes.
PII and access
- Tag PII (
pii: true); restrict views containing it; prefer derived, non-identifying features (hashed or aggregated) over raw identifiers. - Do not put direct identifiers (email, phone, name) into model features at all; keep them in the entity key mapping service.
- Record lawful basis/purpose for sensitive features (regulated domains).
Versioning and change management
| Change | Process |
|---|---|
| New feature | add to a view; existing services unaffected |
| Definition change | new feature name (spend_30d_v2) or new view; never silently change meaning |
| Removal | deprecate (tag + date), notify service owners, remove after consumers migrate |
| Model update | new feature service model_v2; keep v1 until the old model is retired |
Review checklist
- Entities use stable business keys; composite keys where needed.
- Every source has an availability-correct timestamp and a created timestamp if rows are corrected.
- Views split by entity, source and cadence; < ~50 features each.
- TTL >= cadence + delay; freshness SLA for online views; monitoring in place.
- Names follow the conventions; every aggregate has a window.
- Every feature and view has a description and an owner.
- PII tagged and access-controlled; no labels in feature views.
- Feature services per model version.
- Historical retrieval asserts row counts; online/offline parity check scheduled.
anomaly-detection-model-builder
Build and evaluate anomaly detection for tabular, sensor and transaction data with few or no labels: define the anomaly type (point, contextual, collective), choose detectors (robust z-score/MAD and IQR rules, Mahalanobis with robust covariance, k-NN distance, LOF, Isolation Forest, one-class SVM, PCA reconstruction / autoencoders, forecasting residuals for time series), combine them by rank, explain each alert by its deviating features, set thresholds from an alert budget or the score distribution, and evaluate with precision@k, PR-AUC and recall on known incidents, or detector agreement when no labels exist. Includes pure-Python detectors (robust z, Mahalanobis, kNN, Isolation Forest, ensemble) and an evaluation script.
catboost-model-builder
Build CatBoost models that use its strengths correctly: raw categorical columns as cat_features (ordered target statistics, no manual encoding), text_features, Pool construction with NaN-safe categories, ordered boosting and has_time for temporal data, symmetric (oblivious) trees, overfitting detector with eval_set and use_best_model, losses for classification, regression (Quantile, Poisson, Tweedie) and ranking (YetiRank), class weights, monotone constraints, feature importance (PredictionValuesChange, LossFunctionChange) and SHAP values, GPU training and model export. Plans cat/text/numeric columns from a CSV and generates a runnable training script, and checks parameter sets for the conflicts CatBoost rejects (iterations vs n_estimators, Bayesian bootstrap with subsample, multiple class-weight options, depth > 16, use_best_model without eval_set).
categorical-encoding-selector
Choose and implement the right encoding for every categorical column given the model family and the data: one-hot (with infrequent-level grouping and handle_unknown), ordinal with an explicit order, cross-fitted target encoding (scikit-learn TargetEncoder, CatBoost ordered statistics, category_encoders inside CV), frequency/count encoding, feature hashing, native categorical support in LightGBM, HistGradientBoosting, XGBoost and CatBoost, prefix levels for codes (zip3, NAICS-2), and embeddings. Measures cardinality, rare-level share, unseen-level rate on a time-ordered holdout and target signal per column, emits the ColumnTransformer, and quantifies target-encoding leakage (in-sample vs out-of-fold vs holdout AUC).
classification-model-selector
Choose the right classification algorithm for tabular, text or mixed data and prove the choice with a fair bake-off: logistic regression (L2, L1, elastic net), linear SVM, naive Bayes, k-NN, RBF SVM, decision trees, random forest, extra trees, HistGradientBoosting, LightGBM, XGBoost, CatBoost, explainable boosting machines (EBM/GAM), points scorecards and MLPs, matched to data size, width, categorical and high-cardinality share, missingness, class imbalance, text, and constraints such as interpretability, monotonic constraints, calibrated probabilities, latency budgets and tiny deployable models. Ranks candidates with starting hyperparameters and cautions, then generates a bake-off script running every candidate on identical CV folds with per-fold metrics, fit time and prediction latency.
clustering-algorithm-selector
Choose the right clustering algorithm and check that clustering makes sense at all: cluster tendency (Hopkins statistic with a correlated-features caveat, nearest-neighbour isolation), then k-means / MiniBatchKMeans, Gaussian mixtures with BIC, agglomerative (Ward, average, complete, single) with dendrograms, DBSCAN, HDBSCAN, OPTICS, spectral clustering, BIRCH, mean shift, k-prototypes and k-modes for mixed or categorical data, and Gower distance, matched to data size, expected shapes, noise, varying density, known k, soft memberships, hierarchy and streaming. Covers preprocessing (scaling, encoding, dimensionality, distance metrics), validation (internal indices, bootstrap stability, external labels, business profiling) and assigning new points.
collaborative-filtering-implementer
Implement collaborative filtering recommenders that beat popularity: item-based kNN with cosine, Jaccard or asymmetric similarity, shrinkage, neighbourhood size, positive-only or mean-centred rating signals and 'because you liked X' explanations; implicit ALS with confidence weighting, BPR pairwise ranking and the closed-form EASE model in plain NumPy, tuned on a temporal validation split by recall@k and NDCG@k and refit on all data; plus user-kNN, explicit matrix factorisation, the implicit/LightFM libraries, hyperparameter ranges, cold-start fallbacks, scaling with sparse matrices and approximate nearest neighbours. Produces top-k lists in the format of the recommendation evaluator.