The sovereign AI platform

OWN YOUR
DATA.LOGIC.DECISIONS.

The sovereign AI platform that turns your business into an ontology you own, then runs AI inside your rules. Self-hosted or cloud. Every decision shows its work.

Own your data
Connect everything
Integrate every source into one ontology, a living digital twin of your business that the people who run it can actually read.
Own your logic
Rules you control
Your business logic is written as explicit rules the AI must follow. Your team owns them and can change them in minutes, no redeploy.
Own your decisions
Shows its work
Every decision comes with the rules it followed, the data it used, and the reason behind it. Nothing hidden, nothing you cannot check.
01

Provable, not just explainable

Every decision emits the rules that fired, the data it used, and a tamper-proof record. The reasoning itself, not a log written after the fact.

02

Rules bound the AI

Business logic is written as explicit Prolog rules. When the model and the rules disagree, the rules win. No silent drift, no black box.

03

Composable, not monolithic

Run the full stack or interface with what you already have, using only the parts you need. Verticals like Harmonys ship ready to instantiate, not rebuild.

04

Sovereign by design

Bring your own model, a small one you self-host, or a frontier cloud model when you want it. Run where your data lives: cloud, on-premise, or fully air-gapped.

LOGIC.
FIRST.
ALWAYS.

At the core is Ontology Weaver (OntoWeaver), a proof engine that combines Prolog for logic, language models for reading, and machine learning for prediction. The rules set the boundaries; the AI works inside them. It closes the loop from data to decision to action, and writes the result back to your systems. Every logic path is traceable.


Ontology, a digital twin

Your business as objects

Model your organization as objects, links and actions, a living digital twin, not rows and columns. Readable by the people who run it.

INGEST: structured, unstructured, streaming
REASON: causal chains, constraints
DECIDE: decision + rationale + confidence
ACT: write the decision back to your systems
Ontology Weaver (OntoWeaver)

Prolog + LLM + ML

Symbolic rules bound the language model and the predictions. Every inference is traceable; every decision carries a tamper-proof record.

Prebuilt verticals

Harmonys ships first

Healthcare operations: claims, providers, pre-authorization, fraud, settlement. A working vertical to instantiate, not a framework to assemble.

Deployment

Cloud, on-prem, air-gapped

Runs on infrastructure you control, on the model you choose: small and self-hosted, or frontier in the cloud. No vendor lock-in.

Scenario simulation

Ask "what if?"

Test a decision against the digital twin before you make it. Simulate future states and see the downstream consequences.

PROVABLE,
NOT JUST
EXPLAINABLE.

High-stakes work needs AI that can be questioned, explained, challenged, and overridden. Explainable means you can describe a decision. Provable means you can verify it. That is the gap AmmaLogic is built to close.

ProvableBounded AIDecision provenanceSelf-hostedOn-premAir-gapped
01

Every decision is provable

Not an audit log written after the fact, but the reasoning itself. The rules that fired, the data used, the AI's suggestion, the final outcome, and a tamper-proof record tying them together.

02

The AI stays inside the rules

Prolog constraints bound the model. When the AI and the rules disagree, the rules win. Your team owns the rules and can change them in minutes, no redeploy.

03

Your logic, your model, your building

A small model you self-host, or a frontier model in the cloud. Your call, on-premise or air-gapped when it matters. Sovereignty by architecture, not by promise. Nothing travels anywhere you don't allow.

BUILT FOR
HIGH-STAKES
SECTORS.

AmmaLogic is built for the industries where a wrong decision carries real consequences. Healthcare comes first, the rest follow the same model.

HC

Healthcare

The first vertical. Harmonys models claims, provider networks, pre-authorization, fraud and settlement, in active development.

FS

Financial Services

Credit decisioning, fraud and risk, decisions that must be explained and defended. On the roadmap.

GV

Government

Mission analytics and resource allocation, air-gapped where required. On the roadmap.

MF

Manufacturing

Predictive maintenance and quality control, inference at the operational edge. On the roadmap.

LS

Logistics

Routing, demand forecasting and carrier optimization, with a traceable reason for every call. On the roadmap.

YOUR
INFRA.
OUR
LOGIC.

Three deployment modes. One platform. Small or frontier models, zero vendor lock-in.

Mode 01

Cloud

Multi-region deployment on the cloud of your choice. Managed updates and scaling, without giving up ownership of your data.

Mode 02

On-Premise

Full deployment inside your own data center. Your hardware, your network perimeter, complete data sovereignty.

Mode 03

Air-Gapped

Zero external connectivity, self-hosted models only. For the most sensitive, regulated environments, nothing leaves your perimeter.

CONNECT
YOUR
STACK.

AmmaLogic plugs into the data architecture you already run. No rip-and-replace, no migration risk. Reasoning layered on top of the systems you have.

CRM
ERP
Data Warehouse
Data Lake
BI Tools
Cloud Storage
Object Store
SQL Databases
NoSQL
Streaming
Message Queues
ETL / ELT
Identity / SSO
Observability
Ticketing
REST / Webhooks

THE IDEAS
BEHIND
AMMALOGIC.

The thinking AmmaLogic is built on. Three fields, one goal: decisions you can trust and question.

DE

Decision Engineering

Turning messy choices into clear, traceable reasoning. Built on decision analysis (Ron Howard) and decision intelligence (Lorien Pratt).

ES

Expert Systems

Encoding real know-how as rules a machine can apply. From Edward Feigenbaum's early systems to Douglas Lenat's Cyc and its microtheories.

NS

Neuro-Symbolic AI

Joining neural networks with symbolic logic, so AI can predict and prove. Gary Marcus, Henry Kautz, and Judea Pearl on cause and effect.

AMMALOGIC,
EXPLAINED.

The short version, in plain words. What AmmaLogic is, and how it works.

01

Your business as a map

We turn your business into a map of things and how they connect: patients, claims, providers, orders. This map is the ontology, a living picture of how your organization really works.

02

Rules you can read

The rules that run your business are written in plain, explicit logic, not buried in code. Your team can read them and change them in minutes.

03

AI that stays in bounds

The AI reads and predicts, but the rules set the limits. If the AI and a rule disagree, the rule wins. You get the help without losing control.

04

Proof for every decision

Every decision comes with its receipt: which rules fired, what data was used, what the AI said, and the final answer. You can always check why.

05

Your data, your model, your place

Run a small model on your own machines, or a frontier cloud model when you want more power. Cloud, your own servers, or fully offline. You choose.

NEWS
FROM
AMMALOGIC.

Announcements, product updates, and milestones. Real posts coming soon; the cards below show the layout.

PR

Placeholder: announcement headline goes here

Template card. Replace with a real post: date, a short summary, and a link.

PP

Placeholder: product update goes here

Template card showing the newsroom layout. Nothing here is real yet.

FN

Placeholder: milestone or note goes here

Template card. Add real news as it happens.

START WITH
PROOF.

Speak with a solutions engineer. A real conversation about your environment, your risk tolerance, and what it takes to run AI your teams can actually rely on.

Provable decisionsPrebuilt healthcare verticalSelf-hosted modelsOn-prem or air-gappedNo vendor lock-in