Data
Data Science Studio
Ask a business question; it picks, trains and explains the right model.
$25.00 / $0.50 per model built (up to 1M rows) / 1,000 predictions. One-time setup $9,440, 25% off with another product
What it does
Natural question decomposition
Translates business questions into structured specifications: target variable, unit of analysis, observation horizon, and relevant filters.
Rule-based family mapping
Deterministic decision tables map goals to task families (classification, forecasting, regression, survival, segmentation, anomaly detection).
Rigorous bake-off
Trains candidate models with honest validation (time-ordered splits for temporal data). Simpler models win if within 1% of the top score.
Plain-English driver summaries
Explains predictions in business terms with SHAP-based feature importance, showing leadership exactly what drives the score.
Reusable local prediction API
Serves winning models as a private, versioned local REST endpoint counted under your predictions meter.
Model catalog and algorithms
- Classification: Logistic regression, Random forest, Gradient boosting (XGBoost, LightGBM, CatBoost)
- Regression: Ridge, Lasso, Random forest, Gradient boosting
- Forecasting: Seasonal-naive, Exponential smoothing (Holt-Winters), ARIMA/SARIMAX, Prophet-style
- Price analysis: Log-log regression, Demand curve optimization, Double machine learning
- Customer segmentation: RFM scoring, K-means, Gaussian mixture models
- Anomaly detection: Isolation forest, Robust z-score, Residual decomposition
- Survival & churn timing: Kaplan-Meier, Cox proportional hazards, Boosted survival trees
- Market basket: Association rules (Apriori, FP-growth), Collaborative filtering (ALS)
Guardrails and safety controls
- Automated leakage detection rejects columns populated after the outcome occurred
- Minimum sample thresholds enforce at least 200 observations per outcome class
- Simpler models take precedence over complex ensembles when performance is within 1%
- Every trained model records an immutable snapshot of training data versions and hyperparameters
Where it runs
Your servers
Runs directly on hardware in your datacenter. Operates offline.
Your private cloud
Deployed inside your cloud VPC under your direct administration.
Dedicated private hosting
Single-tenant dedicated instance operated solely for your firm.