CDE · Causal Dynamics Engine
CDE discovers causal structure, governing equations, and directed mechanisms from observational and experimental data across science and engineering.
Featured case studies
Each example shows how CDE discovers directed mechanisms, governing equations, and reviewable causal claims from real scientific and engineering data.
High path fidelity
Tokamak confinement scaling involves coupled operating regimes that correlation-based analyses conflate, producing scaling laws that fail under extrapolation. CDE separated these regimes and produced auditable causal claims reviewable edge by edge, each traced to its supporting data partition and negative-control outcome.
Temporal causal links validated
Climate teleconnections span decades across coupled ocean-atmosphere systems where standard time-series analyses conflate lagged correlations with genuine causal pathways. CDE discovered directed causal links among ocean temperature, atmospheric circulation, and precipitation from long-trajectory satellite and reanalysis data. Negative controls confirmed the identified edges reflected physical coupling, not shared spectral structure.
Near-perfect in-distribution fit
Extracting governing equations from noisy sensor traces typically requires a physicist to hypothesize a model family, fit parameters, and iterate. CDE discovered the harmonic structure of a driven spring-mass system, matching the known analytical form with strong in-distribution fit and stable out-of-sample behavior under Truth Dial validation.
Validated on single-cell data
Predicting cellular responses to unseen genetic perturbations requires causal mechanisms in high-dimensional expression space. CDE learned representations from single-cell perturbation readouts and generalized to held-out combinations beyond the training hull, with causal claims linked to data snapshots and control results in the evidence ledger.
Conservation validated
Chemical reactors generate continuous telemetry, but governing rate laws and transport equations are buried under sensor noise and process variability. CDE extracted interpretable rate laws, mass-balance relationships, and directed causal structure verifiable against known chemistry. Conservation violations caught during negative controls blocked spurious claims from the evidence ledger.
Regime structure discovered
Standard factor models treat market dynamics as stationary, missing regime-dependent volatility structures that drive tail risk. CDE discovered governing expressions and causal dependencies from multi-asset time-series data, capturing dynamics invisible to conventional approaches. Distinct market regimes were identified with causal edges between factors, with Truth Dial at Validate level ensuring discovered boundaries survived permutation and holdout controls.
Process equation extracted
Manufacturing control loops rely on empirical tables maintained manually for years. When conditions shift, operators retune by trial and error. CDE distilled causal structure from high-frequency sensor telemetry, relating actuator states to downstream quality metrics with governed, reproducible process models. Conservation checks validated mass and energy balance consistency.
Causal dynamics
CDE discovers directed causal structure, governing equations, and system mechanisms, selecting and combining internal methods automatically based on the data and the causal question.
Discover what drives what: directed edges with confidence estimates across time-resolved systems.
Extract the equations behind the causal structure: compact laws, regime-dependent dynamics, and conservation relationships grounded in mechanism.
Answer what would change under intervention and which experiment would resolve causal ambiguity fastest. Active experiment design built into the engine.
Automated negative controls applied to every causal claim before promotion.
Causal Graphs
Governing Equations
Intervention Design
Causal Validation
Proof and review
Once CDE surfaces a candidate finding, validation and provenance make it reviewable. Teams can see the data snapshot, the route taken through the workflow, and the control results that support or reject the output.
The evidence ledger ties inputs, intermediate results, controls, and promoted findings together. That matters when teams need to compare runs, justify decisions, or hand results to scientific reviewers and operators.
Controls stress-test discovered structure before it reaches decision-makers. Findings that fail those checks stay visible as failed attempts instead of quietly becoming part of the story.
By domain
Different industries face different causal questions, but the engine is the same: discover directed mechanisms, governing equations, and causal structure from the data teams already collect.
Causal target identification, dose–response mechanisms, and perturbation response modeling from omics data. CDE discovers directed causal structure in single-cell and combinatorial datasets where correlation-based analysis cannot distinguish cause from effect.
Reactor dynamics, grid stability modeling, and renewable output dependencies from sensor telemetry. CDE discovers directed causal structure in time-series data where correlation-based methods confuse cause and effect.
Causal structure–property relationships, phase diagrams, and degradation mechanisms from experimental synthesis data. CDE discovers the directed dependencies between composition, processing conditions, and material performance.
Causal root-cause analysis, process optimization, and quality mechanisms from production-line telemetry. CDE discovers the directed causal chain from process parameters to quality outcomes, with governed outputs that integrate into quality-management systems.
Causal risk factor discovery, regime detection, and directed mechanism discovery from tick and fundamental data. CDE separates causal drivers from correlated noise, with the evidence ledger providing the audit trail compliance teams require.
Climate modeling, pollutant transport, and ecosystem dynamics from satellite and sensor networks. CDE discovers directed temporal causal structure across long trajectories that snapshot correlation analyses miss.
Causal treatment response modeling, biomarker mechanism discovery, and patient-trajectory analysis from clinical and real-world data. CDE discovers directed causal pathways with governance controls that meet traceability standards for clinical decision support.
Semiconductor process characterization, network behavior modeling, and causal performance dependencies from engineering telemetry. CDE produces typed causal claims and reproducible runs that integrate into CI/CD-style research pipelines.
Outputs
CDE does not produce unstructured text or unlabeled predictions. Every discovery run outputs typed scientific claims with metadata linking each one to the data snapshot, discovery mode, governance settings, and validation results that produced it.
This typing matters because downstream consumers — reviewers, regulators, publication workflows, other software systems — need to know what kind of claim they are looking at and how it was produced. Different discovery modes carry different semantics and validation requirements. The type system makes that distinction machine-readable and auditable.
Industries
Industry pages with sector-specific discovery context and application detail.
Start with a causal discovery campaign where directed mechanisms, governing equations, or causal structure will make the difference.