SDX DEFENSE Discuss your first use case
Deterministic defense intelligence

Sensor fusion is not evidence fusion.

Defense intelligence needs more than interpretation.
It needs an executable derivation.

You own the sensors. You fuse the data. SDX derives the finding through explicit mathematics and deterministic execution.

Intelligence findings. Calculated, not composed.
Explicit mathematics. Versioned rules. Deterministic execution. SDX Defense does not ask a language model what to conclude. It executes the method that derives the finding.

Same evidence.Same mandate.Same rules.Same execution.Same analytical finding.
Keep your sensors. Keep your AI. Keep your experts.Add a deterministic core.You engineered the sensors. Who engineered the evidence chain?We own that step.
Sovereign by architecturePurely European.Executed locally.No external connections.
02 / 12Position

You already built everything up to here.

ObservationRF · SAR · EO · IR · Radar · UAS · OSINT
FusionCommon operational picture
AnalyticsPatterns, detections, predictions
EvidenceWhat the combined evidence supports — and what it does notSDX Defense
DecisionHuman and institutional authority

Four of these five layers exist in every serious defense architecture. The fourth is assembled in the briefing — by hand, by experience, and differently each time. That is the step we build.

For armed forces, intelligence agencies and the defense industry: certainty about what your intelligence actually supports — and new findings that come from that certainty. Each with its derivation and its limits.

03 / 12Use cases

Use cases.Nine questions. One executable derivation.

Every requirement reaches someone who has to answer it — and defend the answer when it is challenged. These are the shapes SDX Defense is built to take: what is tasked, and what reaches the brief. Your sensors, your analytics and your analysts keep their work. The derivation is the step that is added.

Adversary activity
Readiness under contest
Attribution & verification
Use case 01 / 09Adversary activity

Drone reconnaissance

“Does the collect support the conclusion that this position was reoccupied — and what else would produce the same signature?”

What SDX contributes

Released UAS collect evaluated against a defined requirement, drawing on further suitable evidence where available. Delivered: a bounded finding on what changed, with its derivation and the explanations it leaves open.

Outside the scope

SDX does not operate or task drones.

01 / 09
04 / 12Analysis

Two analysts. Same data. Different assessments.

Show us what changed.

The question?
The assumptions?
The analytical model?

Different judgments can be legitimate. Their foundations should not be invisible.

SDX makes the method explicit before execution. Experts define and challenge the assumptions. The pipeline executes the agreed calculation. Changing the operator does not change the computed finding.

Change an assumption?Version it.
Change a rule?Record it.
Challenge a finding?Reproduce its derivation.

We do not force experts to agree.

We make their disagreement inspectable.

05 / 12Derivation

An explanation can persuade.

A derivation can be rerun.

SDX Defense does not generate an assessment and attach a justification afterward.

It produces the finding through the evidence chain itself.

01A bounded question.
02Defined variables.
03Explicit mathematical relationships.
04Versioned dependencies.
05Deterministic calculation.
06Verification.
07A sealed result with its permitted claims and limitations.

Relationships across observations can establish findings that no individual source supports alone. Contradictions can weaken an apparently convincing explanation. Unresolved dependencies remain unresolved—not hidden inside a confident narrative.

New findings through explicit mathematics.

Before the decision. Not as justification after it.

The evidence chain is not documentation surrounding the product.

It is the product’s computational foundation.

06 / 12Assessment

What a finding looks like.

With its limits attached.

A modelled scenario:

Explicit assumptions and versioned rules connect observations from five source families to a finding. Competing explanations are evaluated against the same evidence, under the same rules, in the same run.

What the evidence supports, and where that support ends, comes from that one defined evaluation. Both are results of the calculation, not judgements added to it.

Intelligence questionIs the observed activity consistent with sustained subsurface construction?
EOPersistent construction-related surface change
ThermalPersistent thermal anomaly
SARSurface-change pattern
SIGINTChange in emission pattern
InfrastructureSupporting logistics change
H1Sustained subsurface constructionSupported
H2Conventional surface constructionPartially supported
H3Temporary logistics activityInsufficiently supported
17Evidence dependencies
5Source families
2Contradictory observations
3Unresolved dependencies
Authorized assessmentThe combined evidence is consistent with sustained subsurface construction activity within the bounded area and observation period.
Not established by this evidenceMilitary purpose · Specific facility type · Operational status · Intent
Modelled scenario · synthetic values · not a run output

A reproducible score is not a reproducible derivation.

The report presents the result. The deterministic evidence chain produces it. Change the operator, the wording, the briefing — the calculation does not move.

Do not only read the assessment. Reproduce its derivation.

07 / 12Artificial intelligence

Your models were never the weak point.

They were asked to carry a conclusion they cannot derive.

“The model says so” is not an evidence chain. Neither is “our analyst believes so.”

Your existing models can detect patterns and supply observations. Their outputs enter the evidence process with their provenance, assumptions, and limitations—not as unquestionable facts.

SDX’s authoritative core is deterministic. It evaluates defined inputs through explicit, version-controlled mathematics.

A language model may express a verified finding.
It does not determine the finding.

A better-written sentence cannot strengthen weak evidence. A new prompt cannot change the result of a sealed calculation.

The wording may vary.

The evidentiary basis must not.

A briefing. One slide. Someone asks why it is that number and not another.

“That is what the model returned” repeats the result. It does not show how the result came about.

The number made it into the room. Did its derivation?

87% confident.Why?
What does 87% actually measure?
Which source moved the assessment?
Which assumptions and dependencies mattered?
Which competing hypothesis failed?
What remains unresolved?
Which input and rule versions produced it?
The percentage is not the problem. The missing derivation is.

With SDX Defense, the derivation is part of the result.

The finding is computed through explicit mathematics and deterministic execution. Its evidence chain names the inputs, the assumptions and the versioned rules it used — and the limits of the claim that follows from them. Not an explanation assembled once someone asks, but the record the calculation produced on its way to the finding.

The derivation does not stay behind in the system. It comes with the finding.

08 / 12Architecture

You have engineered the defense architecture.

Engineer what can be concluded from it.

For defense institutions, allied programs, and the industrial systems behind them, SDX provides an additional capability—not another replacement platform.

01You provide the mandate. The question to be evidenced, the area, the period, the point of intervention.
02SDX sources the data. Copernicus, climate and land-use records; licensed commercial imagery where the question needs it. Where the question rests on your own collection, it arrives by agreed handover — never by connection.
03The run is built and sealed inside our own controlled environment. No cloud, no customer login, no connection to your systems.
04Nothing leaves but the report.

The layer is conceptual, not an installation. There is no interface to us, and there will not be one — no API, no adapter, no network path. That is not a stage we have yet to reach. It is the design, it is permanent, and it is the reason the evidence core can stay sealed at all. You integrate SDX into your decisions and your process, never into your systems. Preserve your sensors, your processing infrastructure and your specialist expertise. Retain human decision authority. You adopt nothing, install nothing and accredit nothing. The engine itself is never delivered. What reaches you is a sealed evidentiary report with its Evidence Class — and, on request, a controlled review of the underlying evidence, on site and under confidentiality.

Turn a defined analytical method into a repeatable system capability—not a process that has to be reconstructed for every briefing.

Your architecture gains a deterministic evidence foundation.

Your teams gain a common calculation they can inspect and challenge — against the sealed record, under controlled review.

Not another system competing for control.

A separate capability inside the process you already control.

Dark data center aisle lined with illuminated server racks
09 / 12Sovereignty

European by jurisdiction. Local by execution.

Nothing connects outward.

The evidence core runs on European hardware we operate ourselves, through a deterministic pipeline, with no path to an external system. A local language model puts verified findings into words. It does not calculate them, and it is not reachable from outside.

01European company. European data center.
02No connection to external systems or services.
03No cloud. No external AI service.
04Local language model. Wording only.

There is no third-country transfer, because there is no transfer at all. No foreign jurisdiction reaches a system that holds no outbound connection — no Standard Contractual Clauses to negotiate, no adequacy decision to rely on.

Processing happens in our own rooms, with no third-party access, to anyone. Data arrives and leaves by agreed handover, never by connection. This is an architectural property, not a certificate. We claim no physical air gap and no accreditation. What we claim is testable: name the connection you expect to find, and we will show you it does not exist.

10 / 12Test & evaluation

You would test the sensor.

Test the inference.

Repeatability is engineered. Accuracy is tested.

Deterministic execution does not make a flawed assumption true. The inputs are fixed and agreed before the run, and the assumptions are declared rather than absorbed. Reproducibility then proves one thing exactly: that this finding follows from those inputs. It does not prove the inputs. That is what the reference test is for.

01Fix the inputs.
02Version the method.
03Reproduce the finding.
04Test it against independent reference evidence.

An evidence score is not automatically a probability. A successful demonstration is not validation.

No certainty by presentation.

No authority by algorithm.

If the claimed analytical contribution does not survive the test, it does not belong in the claim.

11 / 12What you receive

Nothing executable leaves our rooms.

A report you can take apart.

The engine stays where it runs. The derivation travels.

What you receive is the finding and the chain that produced it, as a document. Not software, not a model, not a method you have to take on trust.

01A deterministically generated report. Same mandate, same inputs, same document.
02A document identity and an evidence class, carried on the report itself.
03Every figure traceable to the sealed data it came from. We retain that data for ten years.
04On request, a controlled review of the underlying evidence — a named reviewer, on site, under confidentiality.

The report shows what the finding rests on. It does not show how the engine is built — no rule libraries, no implementation, no internals.

The finding carries your signature, not ours. What we warrant is that it can be reproduced and checked. Not that the assumptions you agreed to were the right ones.

12 / 12Engagement

Where is the deterministic evidence core in your defense architecture?

Not the system that collects the observations.

Not the model that identifies a pattern.

Not the briefing that presents an interpretation.

The executable chain that establishes how the finding follows—and lets another team reproduce it.

That is the conversation.

Bring one unclassified intelligence question. Start with an abstract description, not a sensitive dataset. Define where a deterministic evidence layer could contribute and what a credible test must demonstrate.

What a first mandate asks of you
01The question — what is to be evidenced.
02The area, the period, and the point of intervention it applies to.
03Confirmation of the test criteria, fixed before the run.
04No data and no systems. SDX sources what the question needs.

A mandate is a question, not a licence. Nothing executable is delivered, so there is nothing to install, accredit or maintain on your side.

We have no deployed references. You would be the first, and we would rather say that here than have you discover it in the second meeting. What we can offer instead is a run you can test against evidence we never saw.

You do not need another promise of intelligence.

You need a derivation you can put to the test.

The next decade of defense intelligence will not be decided by who collects the most.

It will be decided by who can prove what it means.

Discuss your first deterministic evidence use case.

Request a conversation

Legal Notice / Privacy

1. Data controller and website operator
SDX Defense is operated by Scalarfield Dynamics S.r.l. (in formation).

Scalarfield Dynamics S.r.l.
Via San Martino 72
84050 Laureana Cilento, Italy

Represented by its shareholders:
Marcello Raffaele Avagliano
Prof. Dr. Sascha Kreiskott

Method and scientific review

Marcello Raffaele Avagliano — Founder, Managing Director and Method Architect
Conceived and developed the SDX methodology.
Deterministic mathematical modelling · Evidence-chain architecture · Algorithm design · Statistical inference · Reproducible analytical methods · Software architecture · Institutional governance

Prof. Dr. Sascha Kreiskott — Co-Founder and Scientific Reviewer
Reviews and challenges assumptions, methods and findings.
Scientific review · Methodological challenge · Scientific reproducibility · Experimental physics · Data protection · Technology governance

Contact: info@sdx-defense.com

2. Categories of personal data processed
a) Server log data: IP address, browser user-agent, timestamp, requested resource. Legal basis: Art. 6(1)(f) GDPR (legitimate interest in operating and securing the website). Retention: maximum 30 days, then deletion or aggregation.
b) Email correspondence: any data voluntarily provided when contacting info@sdx-defense.com. Legal basis: Art. 6(1)(b) GDPR (pre-contractual measures) or Art. 6(1)(f) GDPR (legitimate interest in responding to inquiries). Retention: as long as necessary for the inquiry, then archived for 24 months in line with standard business correspondence practice.

3. Recipients
No personal data is shared with third parties except as required by law, as necessary for hosting infrastructure, or as set out in this section. The hosting provider acts as data processor under Art. 28 GDPR.

This website displays a compliance badge whose image file is retrieved from eab-compliance.eu, operated by EAB Compliance (European AI Bureau), established in the European Union. When a page is loaded, your IP address, browser user-agent, referring URL and the time of the request are transmitted to that server and processed there under its own responsibility. Legal basis: Art. 6(1)(f) GDPR (legitimate interest in displaying the compliance programme this website participates in). Apart from this, no content is loaded from third parties.

4. International transfers
No personal data is transferred outside the EEA. Should service providers operate outside the EEA in the future, transfers will occur under Standard Contractual Clauses approved by the European Commission.

5. Cookies
This website does not deploy cookies. No analytics, advertising, or third-party tracking technologies are used. The compliance badge described in section 3 is a static image; it is not embedded for analytics or tracking purposes.

6. Rights of the data subject
Under Art. 15–22 GDPR you have the right to access, rectification, erasure, restriction, data portability, and objection. Requests may be sent to info@sdx-defense.com.

7. Right to lodge a complaint
You have the right to lodge a complaint with the Garante per la Protezione dei Dati Personali (www.garanteprivacy.it).

8. Changes
This policy will be updated as the legal structure of Scalarfield Dynamics S.r.l. evolves.