Build, validate, govern and monitor predictive models and AI agents from one integrated engineering and assurance platform.
Build interpretable ML models and understand where they work, where they fail and how reliable their predictions are.
Model Studio → · Model Assurance →Build governed agents with structured knowledge, deterministic controls and verifiable behavior.
Agent Studio → · Agent Assurance →AI should not be trusted because it produces impressive demonstrations. It should be trusted because its behavior can be measured, challenged, reproduced and evidenced.
Every GeometriX product is a stage of the same engineering discipline, applied to both classes of AI.
The model zoo exposes how each prediction is produced. GAMI-Net, GLM trees, neural trees: architectures you can read before any post-hoc explainer runs.
Accuracy is an average. Assurance finds where AI fails rather than hiding failures inside aggregate metrics.
Agents evaluated against graph-verified evidence, not another model’s opinion.
Critical policies enforced outside probabilistic LLM reasoning. Same input, same decision, every time.
Evidence generated as AI operates — tamper-evident, integrity-verified — not reconstructed afterward.
GeometriX agents drive model building and validation through 24 registered model tools. The two worlds already share one loop.
Banking, insurance, healthcare and other high-consequence environments — where AI decisions must be explainable, testable, reproducible and defensible.
Independent, reproducible testing under one methodology for models and agents.
For MRM teams →Contracts, gated tools and evidence trails from the first prototype to production.
For AI programs →Deterministic enforcement with a queryable evidence trail behind every decision.
Governance →The platform ships as two Python packages you can install now: modeva for predictive AI and knowlytix for agentic AI.
Load data, train an interpretable model, understand it — a data scientist’s everyday loop:
from modeva import DataSet from modeva.models import MoLGBMRegressor ds = DataSet() ds.load("BikeSharing") model = MoLGBMRegressor(name="LGBM", max_depth=2, n_estimators=100) model.fit(ds.train_x, ds.train_y.ravel())
Or tell a governed agent what you want in plain English, and let it do the model work under policy:
from knowlytix.harness import Harness, HarnessConfig h = Harness(HarnessConfig(stores_dir="gms_stores")) h.register_gms_tools() h.register_modeva_tools() response = h.run("Load BikeSharing, train XGB depth-2, " "run FANOVA, diagnose accuracy")
