Foundation
Has completed the module and can explain it.
A formalized skill-set list you can attach to a job title. Seven competencies, three levels, four roles — and every level is backed by evidence in the candidate's own NGARi Corps record, not by a certificate anyone could print.
Framework v1.0 · 2026-10-02 · machine-readable at /api/corps/competencies
Three levels. The top one is not self-serve.
Has completed the module and can explain it.
Has done it on hardware they own, and it is in their record.
NGARi has verified it against the system of record.
Time served, attendance, or a self-declared mastery. A level with no evidence behind it is not awarded, and Verified cannot be reached without a credential NGARi issues against the system of record — so a candidate cannot mark it for themselves.
What the holder can do, and exactly what proves it.
Explain where the model runs, where the data goes and who can switch it off — and tell an air-gapped node apart from a dedicated tenant.
This is the sentence a client will ask you to defend before they sign.
Every pathway
Load a model on your own node, run a metered call, and read the audit entry it wrote.
The difference between showing someone a chatbot and showing them a system.
Every pathway
Show a client, an auditor or a funder that a record cannot be quietly edited, and say plainly what that does and does not prove.
The audit chain is the objection-killer; misquoting it destroys it.
Every pathway
Answer "we can't put our data in the cloud" with a demonstration instead of a promise — and know which three claims you must never make.
Almost every local-AI sale is decided in this one conversation.
Every pathway
Scope a small on-premises deployment, price it against the published ladder, and write down what maintenance includes.
The failure mode is not losing the deal, it is winning an unpriced one.
Every pathway
Own, flash, configure and place a NGARi appliance, and hand it over to someone who will use it without you.
A role that maintains boxes has to start with having stood one up.
Every pathway
Run the four foundations as a cohort, and assess a learner against the same questions a client asks.
This is what makes a trainer rather than an operator.
Pathway B
Job titles, the levels each one needs, and the block you paste into the advert.
Keeps NGARi appliances healthy in the field: provisioned, patched, backed up, documented, and reachable when a site needs help.
Entry role. No credential required.
Day to day:
Paste the block below under the job title in an advert, or fetch it as plain text from /api/corps/competencies?role=ngari-node-maintainer .
NGARi Node Maintainer — competency requirements
NGARi Corps Workforce Competency Framework v1.0 (2026-10-02)
Keeps NGARi appliances healthy in the field: provisioned, patched, backed up, documented, and reachable when a site needs help.
A candidate meets this role when they hold the following levels against the framework. Every level is backed by evidence in the candidate's own record:
C-HW-1 Appliance ownership and provisioning — Practitioner
Own, flash, configure and place a NGARi appliance, and hand it over to someone who will use it without you.
C-OPS-1 Node operations — Practitioner
Load a model on your own node, run a metered call, and read the audit entry it wrote.
C-AUD-1 Audit and evidence — Foundation
Show a client, an auditor or a funder that a record cannot be quietly edited, and say plainly what that does and does not prove.
C-COM-1 Scope, price and support a deployment — Foundation
Scope a small on-premises deployment, price it against the published ladder, and write down what maintenance includes.
Evidence the framework accepts, per level: a timestamped attestation in the candidate's NGARi Corps record, a metered inference call on their own tenant, a paid appliance, or a NGARi-issued credential. Self-declared only, up to Practitioner; Verified requires a NGARi-issued credential.
Day to day:
- Install and provision an appliance at a site, then hand it over.
- Keep the runtime patched and the disk from filling up.
- Verify the configuration backup restores, not just that it exists.
- Read the audit trail when an owner asks what ran on their box.
- Be the named person a site calls when it stops.
Framework and candidate verification: https://ngari.ai/corps/competencies
Machine-readable: https://ngari.ai/api/corps/competencies Runs sovereign AI for a business or a community: the node, the client conversation, and the evidence that backs the claim.
Credential: CSAO (Pathway A). What each credential requires is on the credentials section.
Day to day:
Paste the block below under the job title in an advert, or fetch it as plain text from /api/corps/competencies?role=sovereign-ai-operator .
Certified Sovereign AI Operator — competency requirements
NGARi Corps Workforce Competency Framework v1.0 (2026-10-02)
Runs sovereign AI for a business or a community: the node, the client conversation, and the evidence that backs the claim.
A candidate meets this role when they hold the following levels against the framework. Every level is backed by evidence in the candidate's own record:
C-SOV-1 Sovereign AI foundations — Practitioner
Explain where the model runs, where the data goes and who can switch it off — and tell an air-gapped node apart from a dedicated tenant.
C-OPS-1 Node operations — Practitioner
Load a model on your own node, run a metered call, and read the audit entry it wrote.
C-AUD-1 Audit and evidence — Practitioner
Show a client, an auditor or a funder that a record cannot be quietly edited, and say plainly what that does and does not prove.
C-CLI-1 The data-privacy conversation — Practitioner
Answer "we can't put our data in the cloud" with a demonstration instead of a promise — and know which three claims you must never make.
C-COM-1 Scope, price and support a deployment — Foundation
Scope a small on-premises deployment, price it against the published ladder, and write down what maintenance includes.
Credential: CSAO (Pathway A; see https://ngari.ai/corps/#credentials)
Evidence the framework accepts, per level: a timestamped attestation in the candidate's NGARi Corps record, a metered inference call on their own tenant, a paid appliance, or a NGARi-issued credential. Self-declared only, up to Practitioner; Verified requires a NGARi-issued credential.
Day to day:
- Operate the node and keep a metered, audited record of what it did.
- Answer the data-privacy objection on the device, not on a slide.
- Scope and price a small deployment against the published ladder.
- Escalate to a maintainer when hardware is the problem.
Framework and candidate verification: https://ngari.ai/corps/competencies
Machine-readable: https://ngari.ai/api/corps/competencies Turns the four foundations into operators: runs cohorts, assesses learners, and stands up the first deployment they learn on.
Credential: CSSA (Pathway B). What each credential requires is on the credentials section.
Day to day:
Paste the block below under the job title in an advert, or fetch it as plain text from /api/corps/competencies?role=sovereign-ai-trainer .
Certified Sovereign AI Trainer — competency requirements
NGARi Corps Workforce Competency Framework v1.0 (2026-10-02)
Turns the four foundations into operators: runs cohorts, assesses learners, and stands up the first deployment they learn on.
A candidate meets this role when they hold the following levels against the framework. Every level is backed by evidence in the candidate's own record:
C-SOV-1 Sovereign AI foundations — Practitioner
Explain where the model runs, where the data goes and who can switch it off — and tell an air-gapped node apart from a dedicated tenant.
C-OPS-1 Node operations — Practitioner
Load a model on your own node, run a metered call, and read the audit entry it wrote.
C-AUD-1 Audit and evidence — Practitioner
Show a client, an auditor or a funder that a record cannot be quietly edited, and say plainly what that does and does not prove.
C-CLI-1 The data-privacy conversation — Practitioner
Answer "we can't put our data in the cloud" with a demonstration instead of a promise — and know which three claims you must never make.
C-COM-1 Scope, price and support a deployment — Practitioner
Scope a small on-premises deployment, price it against the published ladder, and write down what maintenance includes.
C-TCH-1 Teach and assess a cohort — Practitioner
Run the four foundations as a cohort, and assess a learner against the same questions a client asks.
Credential: CSSA (Pathway B; see https://ngari.ai/corps/#credentials)
Evidence the framework accepts, per level: a timestamped attestation in the candidate's NGARi Corps record, a metered inference call on their own tenant, a paid appliance, or a NGARi-issued credential. Self-declared only, up to Practitioner; Verified requires a NGARi-issued credential.
Day to day:
- Teach modules 1 to 4 and assess learners on the same questions clients ask.
- Stand up at least one real site on hardware you own.
- Enrol and run a cohort from first session to assessment.
- Keep your own record current, because your learners' credentials depend on it.
Framework and candidate verification: https://ngari.ai/corps/competencies
Machine-readable: https://ngari.ai/api/corps/competencies Owns a place rather than a box: several sites in one region, the relationships, and the reporting a funder or a local government needs.
Day to day:
Paste the block below under the job title in an advert, or fetch it as plain text from /api/corps/competencies?role=community-deployment-lead .
Community Deployment Lead — competency requirements
NGARi Corps Workforce Competency Framework v1.0 (2026-10-02)
Owns a place rather than a box: several sites in one region, the relationships, and the reporting a funder or a local government needs.
A candidate meets this role when they hold the following levels against the framework. Every level is backed by evidence in the candidate's own record:
C-HW-1 Appliance ownership and provisioning — Practitioner
Own, flash, configure and place a NGARi appliance, and hand it over to someone who will use it without you.
C-SOV-1 Sovereign AI foundations — Practitioner
Explain where the model runs, where the data goes and who can switch it off — and tell an air-gapped node apart from a dedicated tenant.
C-CLI-1 The data-privacy conversation — Practitioner
Answer "we can't put our data in the cloud" with a demonstration instead of a promise — and know which three claims you must never make.
C-COM-1 Scope, price and support a deployment — Practitioner
Scope a small on-premises deployment, price it against the published ladder, and write down what maintenance includes.
C-OPS-1 Node operations — Practitioner
Load a model on your own node, run a metered call, and read the audit entry it wrote.
Evidence the framework accepts, per level: a timestamped attestation in the candidate's NGARi Corps record, a metered inference call on their own tenant, a paid appliance, or a NGARi-issued credential. Self-declared only, up to Practitioner; Verified requires a NGARi-issued credential.
Day to day:
- Run several sites in one region and keep them all reachable.
- Translate site problems into a maintenance and replacement plan.
- Report measured usage and outcomes to a funder, from the record.
- Train the next maintainer for the region.
Framework and candidate verification: https://ngari.ai/corps/competencies
Machine-readable: https://ngari.ai/api/corps/competencies Every level above is read from the candidate's own Corps record: a timestamped attestation they wrote, a metered inference call on their own tenant, a paid appliance, or a credential NGARi issued. A credential is signed and anchored to the hardware that earned it, and can be verified without an account at /api/corps/verify/<credential_id>. A member can open their own record — including which competencies they hold and what is still missing for a role — at the workspace.
Verified is reachable on Pathway B, which carries the teaching step and the credential. Pathway A and the Explorer pathway top out at Practitioner — that is a limit of the current programme, stated here rather than left for a candidate to discover. The framework is also deliberately narrow: it covers deployment, operation, audit and teaching of sovereign AI on hardware you own, and makes no claim about machine learning research or model training.
The framework is published so an employer, a funder or a workforce board can adopt a role definition without inventing one. It is versioned (v1.0, 2026-10-02); a level awarded under one version keeps its meaning because the evidence lives in the member's record. To propose a competency or adopt a role, write to [email protected].