GUIDON reaches AWS Marketplace with offline AI for field accountability

The GUIDON, now listed in AWS Marketplace, takes to Amazon's catalog a well-defined proposal: answer questions in natural language about encrypted local data without relying on the cloud when the connection falls. Each query generates an immutable and signed audit record, and the data is at rest under AES-256-GCM encryption. The identity federation goes through OIDC, LDAP, SAML 2.0 and JWT.

Offline AI: What changes in signalless operation?

The inference exits the cloud and goes to the device. In the field — electrical network maintenance, environmental inspection, remote logistics — the technician needs to consult procedures, checklists, and history at exactly the time of execution, and that is where the connection does not exist. Today the alternative is usually the printed manual or the query postponed to the end of the shift when the vehicle returns to the base. GUIDON shortens this cycle by responding on the encrypted local base without round-trip to a cloud region.

Why is the immutable audit the central point of offline AI?

In field operations, accountability is the product. Each question and answer becomes a signed and unchanging record: who consulted what, when and based on which local dataset. The AES-256-GCM is an authenticated cipher — in addition to hiding the contents, it signals tampering in the stored data. For those who need to prove, after the fact, that a decision followed a documented procedure, that is the difference between a log and evidence.

How does the identity federation (OIDC, LDAP, SAML 2.0 and JWT) work on offline AI?

Four protocols cover different scenarios: LDAP for corporate directories, SAML 2.0 and OIDC for identity providers, JWT for token-based systems integrations. The practical effect is not to create a parallel base of credentials within the device — one of the classic weaknesses of offline tools. Access to sensitive data in the field remains subject to company identity policies.

How does GUIDON compare to other offline AI bets?

It does not dispute the same layer of NVIDIA Jetson, which is hardware platform for edge, nor Microduck, US$399, which places an LLM running locally on a dedicated device. GUIDON is positioned above this: it is accountability software for field operation, which can run on existing hardware in the company. It is also different from hybrid arrangements such as Edge Impulse with AWS, aimed at locating objects in real time, and governance initiatives such as LogiRoot, which operates on the evidence trail of corporate AI.

What the offline AI announcement still does not answer — and the weight of it in Brazil

The listing does not detail model size, minimum hardware, query latency, supported langauges or price beyond the marketplace page. For Brazilian teams, this is where the evaluation begins: sectors with long periods without connectivity — energy distribution, mining, sanitation, environmental inspection in the Amazon, agribusiness — are natural candidates. The adoption, however, depends on two factors that the announcement does not solve: the adequacy to LGPD at rest in equipment subject to loss or theft, and the way immutable records are exported to internal audit systems.

Frequently Asked Questions about offline AI and GUIDON

Does GUIDON work without the internet?

Yeah. The proposal is precisely to execute the AI on the device itself, answering questions in natural language about encrypted local data without relying on the cloud during a connection drop.

What security and identity standards does the solution use?

AES-256-GCM encryption at rest for local data and unchangeable and signed audit records, with identity federation by OIDC, LDAP, SAML 2.0 and JWT.

Where is GUIDON available?

The solution is listed in AWS Marketplace, where the product page concentrates the commercial and technical conditions disclosed so far.