Gaze–Gesture Accessibility Interaction Architect — Consumer Devices

The mandate

The device portfolio is combining eye tracking, hand gesture, voice, switches and adaptive prediction so people can move between interaction modes as ability, fatigue and environment change. Current teams validate each modality separately. They cannot yet show how calibration, inferred intent, mode transition, confirmation and recovery behave when one signal weakens or when personal adaptation mistakes involuntary movement for command.

The deliverable is an Adaptive Accessibility Interaction Architecture for communication, navigation and safety-critical account actions. It will define user intent, capability profile, consent, sensor observation, calibration state, modality confidence, fusion, proposed action, confirmation, execution, correction and safe fallback. Three reference journeys will avoid a normative assumption about movement, speech, gaze duration or caregiver participation.

Milestone one at week four supplies participatory research, exclusion risks and consequence-ranked interaction failures. Week nine concludes milestone two with the multimodal state model and privacy boundaries. At week sixteen, milestone three delivers reference journeys and fatigue or sensor-loss trials. The accepted architecture, design rules, test corpus requirements and internal qualification pack complete milestone four at week twenty-two.

Acceptance requires disabled-user and accessibility specialists to reconstruct twelve unseen actions through calibration, modality evidence, confidence and confirmation; an involuntary gaze fixation and mid-task capability change must fail safely; and Privacy must reperform inference minimisation tests. The design chief signs after client teams run a caregiver-transition and low-light scenario without consultant help.

The client will provide consented research evidence, accessibility needs, calibration logs, multimodal interaction traces, error and abandonment studies, hardware constraints, privacy assessments and representative devices. Compensated participants and accessibility researchers will shape tests; client engineers build prototypes. The scope excludes medical inference, biometric identity deployment, production code, product certification and any attempt to diagnose a user’s capability.

Why this is external work

Modality teams optimise their own sensors, while accessibility defects emerge in transitions and in assumptions that typical users never encounter. Internal product incentives can also favour completion rate over autonomy and recoverability. Independent architecture supplies participatory challenge and cross-modal evidence without classifying disability, prescribing an aid or designing biometric surveillance.

What you will own

• Map capability preference, consent, calibration, observation, inferred intent, modality confidence, confirmation, execution, correction and abandonment.
• Define transition rules when fatigue, tremor, speech change, lighting, posture, device position or caregiver presence alters signal usefulness.
• Design proportional confirmation for communication, navigation, purchase, privacy and account actions with different reversal consequences.
• Exercise involuntary fixation, ambiguous gesture, speech loss, switch failure, low light, interrupted calibration and mid-task mode change.
• Specify adaptation transparency so users can inspect, reject, reset or constrain learned interaction without losing essential access.
• Assess sensors and fusion components through inclusion coverage, on-device privacy, latency, calibration burden, portability and supplier exit.
• Transfer participatory test design and adaptive-state qualification to permanent accessibility, design and engineering owners.

Candidate qualifications

• Led multimodal accessibility or assistive-interaction architecture across consumer devices used by people with varied motor and communication needs.
• Combined gaze, gesture, speech, switches or alternative input without allowing one modality’s confidence to erase user intent.
• Designed calibration, confirmation and recovery for fatigue, involuntary movement, environmental change and progressive capability variation.
• Conducted compensated participatory research with disabled users while protecting dignity, consent and sensitive interaction evidence.
• Worked with privacy and product-safety teams without recasting accessibility signals as diagnosis or covert biometric identity.
• Delivered reusable interaction rules and test evidence that internal teams applied after an external design engagement ended.

Non-negotiables

• The named architect must lead Seoul participatory sessions and both low-light and caregiver-transition acceptance exercises.
• No financial relationship may exist with gaze, gesture, speech, identity or accessibility-component suppliers evaluated.
• Disabled participants must be compensated, consented and empowered to stop or withdraw their research evidence.
• Medical inference, biometric identity, production implementation and product certification are expressly excluded.

1. 49 words maximum. Describe a multimodal interaction where a high-confidence signal contradicted the user’s actual intent.
2. 49 words maximum. How did you let a user inspect and reset adaptation without losing essential access?
3. 49 words maximum. Which client evidence is required before testing an involuntary-gaze safety case?