kint

Open Source

Built in the Open

We build core optimization technology as open source. Production-tested, well-documented, free to use.

surrox

Python 3.13+

Blackbox surrogate-based optimization framework. Define your problem declaratively, train ensemble surrogates on historical data, optimize on the learned models, get explained recommendations with uncertainty intervals.

GitHub Docs
$ pip install surrox[all]

Declarative Problem Definition

Define variables, objectives, constraints, and domain knowledge with Pydantic models. Immutable, validated, reproducible.

Ensemble Surrogates

XGBoost, LightGBM, Gaussian Process, and TabICL trained with Optuna HPO. Multiple models, one prediction.

Multi-Objective Optimization

NSGA-II/III via pymoo and TuRBO for trust-region Bayesian optimization. Find Pareto frontiers, not single answers.

Conformal Prediction

Distribution-free uncertainty intervals. Know how confident the model is before you act on its recommendation.

Built-In Explainability

SHAP values, PDP/ICE plots, feature importance, what-if analysis. Every recommendation is traceable.

Scenario Comparison

Optimize across different operating conditions. Compare outcomes side by side before committing.

Extrapolation Detection

k-NN distance flags when new data falls outside the training distribution. Know when to trust the model and when not to.

Quick Start

Define. Train. Optimize. Explain.

from surrox import Problem, Variable, Objective, Constraint, run

problem = Problem(
    variables=[
        Variable("temperature", bounds=(150, 300)),
        Variable("pressure", bounds=(1, 10)),
        Variable("flow_rate", bounds=(50, 200)),
    ],
    objectives=[
        Objective("yield", direction="maximize"),
        Objective("energy_cost", direction="minimize"),
    ],
    constraints=[
        Constraint("purity >= 0.95"),
    ],
)

result = run(problem, data=historical_data)
result.best          # optimal parameters
result.explain()     # SHAP, feature importance
result.what_if(temperature=250)  # scenario testing

Documentation

Learn More

surrox and kint

surrox is the open-source foundation behind kint's blackbox optimization. It handles surrogate training, multi-objective optimization, and explainability. kint adds the full platform on top: natural language input, whitebox solvers, API/OEM delivery, data connectors, and enterprise features.

Use surrox standalone for Python-based optimization. Use kint when you need a production platform with API access and integration support.

Need More Than Open Source?

surrox handles blackbox optimization. kint adds whitebox solvers, natural language input, API delivery, and enterprise support. 30 minutes. We'll show you the difference.

Talk to Us Star on GitHub
Book a Demo