Model Validator
-
IT/Tech
Job Description & How to Apply Below
The Model Validator provides independent oversight of AI risk and performance assessments, reviews issues and determines approval, conditional approval or cancellation. They monitor key AI risk indicators such as drift, bias, safety, security and compliance signals and confirm remediation is completed.
Reports to:
Model Risk Manager
The Model Validator provides independent oversight of AI risk and performance assessments, reviews issues and determines approval, conditional approval or cancellation.
- Independently validate model risk and performance assessments
- Decide approval, conditional approval or cancellation of model use
- Monitor drift, bias, safety, security and compliance indicators
- Review issues raised against models and set remediation conditions
- Confirm remediation actions are completed and effective
- Document validation decisions with traceable evidence
The GRC processes this role owns, touches and depends on.
Ownsaccountable for the process
Touchescontributes or approves
Depends onconsumes its output
How the GRC processes connectThree lines of defence set the frame; the chain shows where each process hands off to the next. Colours follow the Model Validator: navy is owned, blue is touched, grey is depended on.
The business and its control owners. Run the processes, own the controls, fix the issues
Risk, compliance and the domain programmes. Set the frameworks, assess, monitor and advise
Internal audit. Independent assurance to the board and audit committee
Who this role reports to, who reports to them, and the peers they work with most.
Reports toModel Risk Manager
Works closely with- Model Owner
- AI Risk Officer
- AI Governance Lead
- Internal Auditor
The Model Validator is the independent voice on whether a model may be used, and the Model Risk Manager, the AI Risk Officer and examiners expect each decision to rest on evidence and be traceable. Smart Suite gives validation a structured workflow with decision logs and linked issues, so independence is demonstrable, not asserted.
Validate independently with a traceable decisionIndependence documented, not assumed
Clear ownership assigns the model owner, business owner and independent reviewer to each model with documented responsibilities, so your separation from development is on the inventory and visible to anyone who checks.
Exceptions tracked, not waved throughException requests run through the same review workflow as approvals, with the condition recorded, so a model operating outside approved scope is a tracked exception rather than an unwritten understanding.
Conceptual soundness checked in contextConnected model context shows the business processes, applications and datasets a model relies on, so you assess conceptual soundness and intended use against what the model actually touches rather than what the documentation says.
Work with model owners, risk and audit- Issues raised with owners and deadlines attached:
Each issue you raise links to the Model Owner, the mitigation, a deadline and the evidence of resolution, so a remediation condition is a tracked item on their list and a visible one on yours. - Early signals before the incident:
Early risk detection surfaces emerging risks from tracked indicators before they become regulatory or operational incidents, so the AI Risk Officer and you discuss a trend while it is still a trend. - Validation files the Internal Auditor can test:
Testing artifacts, review logs and supporting evidence are stored centrally against each model, so when the Internal Auditor samples a validation they read the file you built rather than a summary you write for them.
- Who approved what, when and why:
Audit trails and version history keep full records of each validation decision, the reviewer and the rationale over time, so an examiner reconstructing an approval finds the reasoning you recorded at the time. - Conformity assessments mapped to the regulation:
Framework and regulatory alignment links models and assessments to laws, regulations and frameworks, so the conformity assessment the EU AI Act requires for a high-risk system is traceable to the requirement it satisfies. - Scope kept to what was approved:
Governance coverage tracking records the approved scope, controls and monitoring cycles for each model, so use outside the approved scope is a visible gap on the record…
(If this job is in fact in your jurisdiction, then you may be using a Proxy or VPN to access this site, and to progress further, you should change your connectivity to another mobile device or PC).
View / Apply for Jobs
Matching My Jurisdiction