Empathetic AI assistants

Medical consultant assistant

A health-support avatar that conducts a structured conversation, reads emotional state while it does so, and produces a record with two parts: the facts a person reported and how they reported them. It integrates uploaded health records for context and flags hesitation or distress for a human to read. It does not diagnose and does not replace a clinician.

• Symptom conversation with emotional-state annotation
• Health record context via AriGraph memory
• Explicit non-diagnostic disclaimer on every response

Fitness and Live Mode coaching

The coaching avatar runs in Live Mode, combining real-time vision, voice and opt-in biometric input so it can distinguish effort from strain and adapt within the session instead of after it. Biometric processing is opt-in and revocable per session.

Nutrition advisor

Camera-based food recognition with nutritional breakdown and meal planning that respects dietary restrictions stored in the user's profile. Advice is general wellbeing guidance, not clinical dietetics.

Education and study companion

An adaptive learning assistant that tracks engagement signals — voice energy, response latency, frustration cues — and shortens or re-paces a session rather than pushing through a collapse in attention. Available to verified students through the student programme.

How the assistants differ from a chatbot

Every assistant shares one emotion-recognition pipeline and one memory layer, so context carries between sessions and between assistants that a user has linked. The avatar's visible expression is driven by the same inferred state as the spoken answer, rendered in WebGL with LipSync and degrading to voice-only where WebGL is unavailable.

• Shared multimodal emotion state across all assistants
• Persistent, inspectable and deletable memory
• Custom assistant development and API integration for partners