Models
Physics-guided where the problem provides structure; testable regardless.
Our approach
The Scientific Loop
A recursive process for discovery. Physical measurement is a future component of the loop, clearly separated from the computational system operating today.
Computational loop activePhysical stages planned
Conceptual visual
Process
Every cycle should make the next prediction better.
The loop is not a linear presentation device. Its purpose is to preserve what was learned, expose disagreement, and change what happens next.
- 01Question
Define a measurable scientific objective.
- 02Model
State assumptions and representations.
- 03Simulate
Run controlled computational evaluations.
- 04Analyze
Extract error, uncertainty, and structure.
- 05Learn
Update the model or search policy.
- 06Compare
Test against credible baselines.
- 07Measure*
Collect calibrated physical evidence.
- 08Test*
Run the selected physical experiment.
* Physical capability planned / being built
How we build reliable science
Methods that make the loop accountable.
The method is only useful when another run can reproduce the inputs, assumptions, comparisons, and uncertainty.
Optimization
Multi-objective search under explicit constraints and equal budgets.
Uncertainty
Track what the model does not know and where evidence has value.
Reproducibility
Version inputs, runs, assumptions, and analysis as one record.
Explainability
Understand why a model or policy recommends the next evaluation.
Deep SIML Labs research principleThe goal is not to be right once.
The goal is to learn faster, with evidence.
Conceptual visual
Operating discipline
Comparison before confidence.
SIML must earn its role through direct comparison against conventional optimization methods. A result is more useful when its assumptions, uncertainty, compute budget, and failure conditions remain visible.
Equal budgetsVersioned evidenceExplicit uncertaintyFailure analysis
Explore the research program