Multimodal Clinical Triage Reliability Leader

The mandate

The clinical-AI reliability director role oversees triage service consistency across spoken narrative, typed symptoms and wound images with partial history. A partner network is preparing regional launches, and operating teams require coordinated protocols that maintain consistency while the organisation separates model behaviour from workflow and escalation design.

The interim must start within three weeks and take the executive reliability seat for twelve months. Recruitment for a permanent director will open after the first two partner sites complete a stable ninety-day surveillance cycle in month six; the chosen successor receives six weeks of joint decision-making before the interim’s authority expires, whether or not wider commercial rollout has resumed.

Handover is complete when modality-specific safety cases define allowed and prohibited uses, triage calibration is evidenced across the agreed symptom and demographic cohorts, missing or contradictory inputs reliably force the correct human pathway, and all live sites use one incident taxonomy with weekly drift review. The successor must chair a simulated deterioration event, sign the residual-hazard register and accept the next two release decisions without support.

The interim may suspend a modality, narrow a pathway, mandate clinician review, quarantine a model version, appoint temporary investigation leads and deploy up to £6 million from the approved remediation reserve. Board approval is required to change clinical risk tolerance, resume a paused regional launch, notify outside parties beyond established protocols or exceed that reserve; permanent hiring and alteration of clinicians’ professional accountability are expressly withheld.

Replacement of the electronic health-record estate, redesign of clinician pay, consumer acquisition strategy and expansion into new countries are not part of the seat. The leader may require interfaces, staffing data or launch assumptions from those areas to test safety, but does not own diagnosis policy, medical-device classification strategy or the commercial roadmap unrelated to triage reliability.

Why this seat is open

The departure removed the only executive who connected model evaluation, clinical workflow and live incident surveillance across every modality. Keeping the role inside the existing product hierarchy would make the same launch pressure that contributed to fragmented overrides responsible for judging remediation. The board wants a time-bounded operator who can impose one evidential standard while a permanent appointment is made deliberately.

What you will own

• Establish a common hazard taxonomy linking modality failure, missing context, cohort effect, clinician interaction and patient consequence to one accountable disposition.
• Decide which symptom pathways may continue, require mandatory human confirmation or remain suspended, documenting the evidence and reversibility of each boundary.
• Rebuild test cohorts around clinically material presentations, skin tones, accents, languages, disabilities, device conditions and incomplete histories rather than headline accuracy alone.
• Calibrate urgency and escalation thresholds separately for text, speech and image inputs, then test discordant combinations that should trigger safe abstention.
• Institute live surveillance that joins model version, modality quality, clinician override, queue pressure, deterioration signal and eventual care outcome without concealing denominator shifts.
• Run partner-site readiness reviews in Manchester and Birmingham, closing local workarounds before authorising any controlled return to service.
• Transfer release records, open hazards, cohort baselines, incident simulations and authority maps through a successor-led committee decision observed by the board chair.

Candidate qualifications

• Held executive accountability for clinical safety or AI reliability in a regulated digital-health, urgent-care or diagnostic service used in live patient pathways.
• Personally paused or restricted an algorithmic workflow after detecting a safety signal and can evidence the investigation, containment decision and monitored restart.
• Evaluated systems combining language, speech and medical imagery, including contradictions and quality degradation between modalities rather than isolated benchmark performance.
• Built cohort-level calibration and surveillance that connected technical outputs to clinician action, escalation delay and clinically meaningful downstream outcomes.
• Directed incident review across product engineers, practising clinicians, human-factors specialists, quality leaders and external care partners with competing interpretations of causality.
• Handed a repaired safety operating model to a permanent leader after demonstrating that local sites could apply it without exceptional executive intervention.