Checklist
Decision and value
- The decision, the acting team and its capacity are named.
- The problem is quantified (volume, cost, rate).
- Value is estimated against the current process, with a ceiling.
Success
- One primary metric at the operating point, on a deployment-like split.
- A numeric target and the baseline's value on the same split.
- Guardrails with limits, including per-segment floors.
- An online confirmation plan (A/B, holdout) with duration.
Data
- Label defined as a query with its window and population.
- Each source has an owner, freshness and availability at prediction time.
- Known leaks and exclusions are listed.
Plan
- Phased approach starting with a baseline.
- In-scope and out-of-scope lists.
- Dated milestones with checkable exit criteria.
- Go/no-go criteria per gate.
Risk and ownership
- ML-specific risks with mitigations.
- Team, FTE and run cost.
- Post-launch owner, monitoring and retraining plan.
- Fairness/regulatory review where decisions affect people.
Red-flag phrases
| Phrase | Question to ask |
|---|---|
| "leverage AI/ML" | Which prediction feeds which decision? |
| "high accuracy" | Which metric, at which operating point, versus what? |
| "real-time" | What latency, and is the data real-time too? |
| "we have all the data" | When is each field written relative to the prediction? |
| "the model will learn" | From which label, observed when? |
| "phase 2: productionise" | Who owns serving, monitoring and retraining? |
| "insights" | What will someone do differently with them? |
Comment templates
- Missing baseline: "What does today's process score on recall@400 for the same holdout? Without it we cannot tell whether the model adds value."
- Metric without operating point: "AUC summarises all thresholds; the team calls 400 a week. Please make recall@400 (or precision@400) the primary metric and keep AUC as secondary."
- No out-of-scope: "Please list what this project will not do; real-time scoring and offer choice seem likely to creep in."
- Milestone without exit criterion: "How will we know 'model ready' is done? Suggest: recall@400 >= 0.45 on the holdout with the CI above the rule."
- Leak risk: "Is
account_statusread as of the scoring date or today? If today, it leaks the outcome." - Feedback loop: "Calling customers changes who churns. Can we keep a random 5% untreated group?"
Sign-off roles
| Role | Signs off on |
|---|---|
| Product/business owner | decision, value, capacity, success criteria |
| ML lead | approach, validation plan, risks |
| Data owner(s) | label definition, source availability |
| Acting team lead | workflow fit, capacity, adoption |
| Model risk / legal (if applicable) | fairness, explainability, regulatory |
| Platform/MLOps | serving, monitoring, on-call |
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.