NSense

A methodology for sensing the unknown

Start with the condition—
not the sensor.

Conventional measurement often begins with a predefined property and a device designed to isolate it. NSense begins with a different question: What do we need to know, and how does that condition reveal itself through physical behavior?

Follow the information ↓
01 The information pathway

A measurement is an inference
grounded in physical interaction.

The desired condition may be inaccessible directly. The task is to identify how it changes an interacting system, preserve the resulting evidence, and build a reliable path from observation to inferred state.

UnknownUnderlying
condition
01Physical
interaction
02Multimodal
observation
03Model and
evidence
InferredEstimated
state

The sensor is not necessarily a single device. It may be the entire interacting system.

02 Four stages of inference

Interrogate. Observe.
Model. Infer.

These stages form an iterative experimental cycle rather than a one-way production line. Evidence discovered during modeling may change what we observe; unexpected observations may require a new mode of interrogation.

01

An information-bearing physical interaction.

Interrogate

Create an interaction that makes the condition observable.

We identify or deliberately create physical interactions whose behavior is influenced by the condition of interest. The objective is not simply to stimulate the system, but to make the hidden state express itself in measurable ways.

Questions we ask
  • What physical properties change with the condition?
  • Which interactions could reveal or amplify those changes?
  • Which boundary conditions must be controlled or varied?
Potential methods

Controlled excitation, perturbation, stimulus design, and analysis of naturally occurring interactions.

02

A calibrated, contextual record of the system response.

Observe

Capture the response without reducing it too soon.

We observe how the interacting system responds across the modalities, locations, timescales, and resolutions that may contain relevant information. Calibration and synchronization preserve relationships that a single isolated signal may miss.

Questions we ask
  • Which observable responses carry information?
  • How do signals relate across time, space, and modality?
  • What variability comes from the condition, environment, or instrument?
Potential methods

Multimodal instrumentation, synchronized acquisition, calibration, signal conditioning, and structured data collection.

03

A testable mapping from observations to candidate states.

Model

Find the relationships that connect response to state.

We combine physical understanding with computational analysis to determine which patterns are meaningful, which are incidental, and how observations relate to the underlying condition. Models are selected for explanatory value, predictive performance, and robustness.

Questions we ask
  • Which relationships are physically plausible and repeatable?
  • What features remain informative across specimens and conditions?
  • Where does the model fail, and how certain is its output?
Potential methods

Physical models, system identification, signal processing, feature discovery, machine learning, and multimodal data fusion.

04

An actionable estimate with defined evidence and limits.

Infer

Convert complex patterns into a reliable measurement.

We apply the model to estimate a property, condition, or emerging change that may not be directly measurable. The result must be accompanied by defined operating limits, uncertainty, and evidence appropriate to the intended decision.

Questions we ask
  • What conclusion does the evidence support?
  • How should uncertainty and confounding factors be represented?
  • What threshold or output would be useful in practice?
Potential methods

State estimation, anomaly detection, probabilistic inference, confidence analysis, and decision-oriented validation.

03 Physics and machine intelligence

Neither physics alone.
Nor a black box.

NSense uses physical understanding to decide what to perturb, what to observe, and which relationships are plausible. Computational methods help discover and combine patterns that are too subtle, distributed, or numerous to interpret independently.

01

Physics frames the search.

Mechanisms and constraints guide experimental design, feature selection, and interpretation.

02

Data reveals the relationships.

Signal processing and machine intelligence identify structure across interacting observations.

03

Validation challenges both.

Models must survive controlled variation, competing explanations, and conditions beyond those used to develop them.

04 What we mean by deterministic information

Physical behavior is structured
by cause and constraint.

“Deterministic” does not mean that every measurement is perfectly certain, noiseless, or completely predictable.

It means that physical interactions are shaped by causal relationships. Changes in material state, geometry, environment, or physiology alter the ways a system can respond—even when those effects are subtle, distributed, and mixed with variability.

The measurement problem is to recover relationships that are sufficiently stable, specific, and validated to support an inference. Uncertainty remains part of the result, not something to conceal.

05 A different starting point

Beyond a better version
of the same sensor.

Some problems require improved sensitivity or selectivity from an established sensor. Others require reconsidering what constitutes the sensor in the first place.

Dimension
Conventional sensing
NSense approach
Starting point
A predefined measurand
The condition that must be understood
Physical scope
An isolated, preferred signal
The larger interacting system
Complexity
Suppressed where possible
Preserved when it carries information
Detection target
A known signature
Known signatures or previously unrecognized patterns
Sensor concept
Primarily a device
Potentially the entire interacting system
06 From hypothesis to measurement architecture

A measurement concept
must earn confidence.

NSense treats validation as part of the sensing architecture. The required evidence depends on the intended use, but the path must expose uncertainty, confounding factors, and operating limits.

  1. 01
    Define the condition

    State what must be inferred, why it matters, and what evidence would constitute a useful measurement.

  2. 02
    Form a physical hypothesis

    Identify the mechanisms by which the condition could alter observable behavior.

  3. 03
    Design controlled experiments

    Vary the condition and relevant confounders while collecting calibrated observations.

  4. 04
    Build and challenge the model

    Test candidate relationships against held-out conditions, competing explanations, and uncertainty.

  5. 05
    Validate the inference

    Determine repeatability, sensitivity, specificity, limits, and generalization appropriate to the application.

  6. 06
    Define the architecture

    Translate the evidence into an instrumentation, computation, and validation plan for continued development.

The output of this process may be a feasibility result, an experimental platform, a prototype measurement system, or a development program—depending on the evidence and the problem.

Begin with the unknown

What condition do you need
to recognize—but cannot
measure today?

Bring us the problem, the physical system, and the decision the measurement must support. We will begin by asking where the information may already exist.

Explore opportunity areas ↘