NHS Partner · λ=10.0 SEOM · Clinical-Grade AI

The companion
every elder deserves.

Fari is designed to deliver clinical-grade health monitoring with the warmth of a trusted friend — combining advanced AI with an unwavering commitment to dignity, privacy, and human authority.

Explore Features →

🏥 NHS Innovation Partner 🛡️ SEOM Ethics λ=10.0 🤖 Jetson Orin NX Edge AI

SEOM Score

98.7

Emotion AI

Active

Vitals

Nominal

Forbes — Top Robotics Innovator Fortune — Best AI Startup to Watch NHS Innovation Accelerator Partner CES Innovation Award IEEE Robotics Summit Speaker UK PM No. 10 Recognition BCG AI Summit Chicago TEDx Featured Forbes — Top Robotics Innovator Fortune — Best AI Startup to Watch NHS Innovation Accelerator Partner CES Innovation Award IEEE Robotics Summit Speaker UK PM No. 10 Recognition BCG AI Summit Chicago TEDx Featured

Fari is in active development · All features, specifications, and performance figures represent design targets · Subject to change

Core Capabilities

Three pillars of
compassionate care.

Fari is being designed to address the three dimensions of eldercare that matter most — clinical health monitoring, emotional wellbeing, and physical safety.

❤️

Health Monitoring

Designed for continuous, non-invasive vital sign tracking — heart rate, SpO₂, respiratory rate, skin temperature, and blood pressure estimation. AI models are being developed to detect subtle health pattern changes before they become emergencies.

🧠

Emotional AI

Our Affective Multi-modal Depression Classifier (AMDC) is designed to analyse facial micro-expressions, vocal prosody, and behavioural patterns to assess emotional wellbeing — enabling proactive companionship and caregiver alerts.

🛡️

Safety & Mobility

Fall detection and prevention using the Human Trajectory Detection system (HTD-IRL). Designed to predict fall risk from gait analysis and environmental scanning, with target alert latency of under 3.2 seconds from detection to caregiver notification.

What Fari Does

Designed to care
in every way.

A selection of Fari's planned capabilities — each governed by our SEOM ethical framework to ensure dignity, privacy, and human authority at every interaction.

Health

Medication Management

Designed to deliver personalised medication reminders with visual and audio cues, confirm intake via computer vision, and alert caregivers to missed doses. SEOM Rule F02 governs all medication interactions.

Emotional AI

Companion Conversations

Natural language interaction designed to reduce loneliness. Fari's Socially-aware Trajectory Understanding Model (STUM) is designed to calibrate conversation style based on resident preferences and emotional state.

Safety

Fall Detection & Prevention

Using mmWave radar and depth cameras, Fari is designed to monitor gait patterns and predict fall risk. The HTD-IRL model uses inverse reinforcement learning to identify pre-fall biomechanical signatures.

Mobility

Autonomous Navigation

WBT1 mobility platform designed for smooth, safe navigation in care environments. Multi-sensor SLAM with LiDAR, depth cameras, and ultrasonic arrays enables obstacle avoidance and room-to-room autonomy.

View All 35+ Planned Features →

Fari in Action

Watch what we're
building.

Our Vimeo showcase brings Fari to life — the team, the technology, and the vision behind InGen Dynamics' eldercare AI companion. Features shown are in active development and subject to change.

InGen Dynamics · Fari Eldercare AI · All features in active development · Not yet commercially available View on Vimeo

Ready to bring Fari to your care facility?

We are actively engaging with care providers, health systems, and research partners. Let's explore how Fari could support your residents and staff.

Schedule a Conversation →

Feature Catalogue

35+ capabilities
designed for care.

Every feature Fari is being designed to deliver — each governed by the SEOM ethical framework. All capabilities represent design targets and are subject to change during development.

All features are in development · Specifications represent design targets

Continuous Vital Signs

Health

Designed to monitor heart rate, SpO₂, respiratory rate, skin temperature, and blood pressure estimation continuously via non-invasive sensors.

SEOM F0114-Sensor Suite

Sleep Quality Analysis

Health

Non-contact sleep monitoring designed to track sleep stages, breathing patterns, and nocturnal movement using mmWave radar and ambient sensors.

mmWave RadarNon-Contact

Health Trend Baselines

Health

AI designed to establish individual baseline health profiles during a 14-day onboarding period, enabling detection of subtle deviations from personal norms.

GRPO ModelPersonalised

Environmental Monitoring

Health

Designed to track room temperature, humidity, air quality, and light levels to identify environmental factors affecting resident health and comfort.

BME680Ambient

EHR Integration

Health

Designed to synchronise observations with Electronic Health Records via HL7 FHIR R4 APIs, enabling seamless clinical data flow with target sync latency under 5 minutes.

FHIR R4HL7

Medication Reminders

Medication

Personalised audio and visual medication reminders designed with escalation protocols — gentle prompt → detailed instructions → caregiver alert. Governed by SEOM Rule F02.

SEOM F023-Tier Escalation

Intake Confirmation

Medication

Computer vision designed to visually confirm medication intake and detect potential missed doses, logging all events for clinical review.

Computer VisionSEOM F02

Emotion Recognition

Emotional AI

AMDC model designed to analyse facial micro-expressions, vocal prosody, and behavioural patterns to assess emotional state with target accuracy of 87.4%.

AMDCMulti-modal

Companion Conversations

Emotional AI

Natural language interaction designed to reduce feelings of loneliness. STUM model calibrates conversation style, topic selection, and emotional tone to each resident.

STUMNLP

Activity Engagement

Emotional AI

Designed to suggest and facilitate cognitive exercises, music therapy, reminiscence activities, and gentle physical exercises tailored to individual capability and preference.

PersonalisedAdaptive

Family Connection

Emotional AI

Video calling with auto-framing (target 1080p), wellbeing digest sharing, and activity photo sharing — all governed by SEOM F08 and F10 consent rules.

SEOM F08SEOM F10

Fall Detection

Safety

Multi-sensor fall detection designed with target response time of under 3.2 seconds from detection to caregiver notification. Uses depth cameras and mmWave radar.

SEOM F03HTD-IRL

Fall Prevention

Safety

HTD-IRL model designed to analyse gait patterns using inverse reinforcement learning to identify pre-fall biomechanical signatures and alert staff proactively.

Gait AnalysisPredictive

Emergency Protocols

Safety

SEOM Rule F03 mandates that emergency detection always takes priority over all other operations. Designed with multi-channel alerting: on-device, dashboard, mobile push, and SMS.

SEOM F03Priority Override

Wandering Detection

Safety

Designed for residents with cognitive impairment — monitors movement patterns and triggers alerts when unusual wandering behaviour is detected, especially during night hours.

Memory CareGeofencing

Autonomous Navigation

Mobility

WBT1 mobility platform with multi-sensor SLAM using LiDAR, depth cameras, and ultrasonic arrays. Designed for smooth, safe movement through complex care environments.

WBT1SLAM

Room-to-Room Patrol

Mobility

Designed for scheduled and on-demand room visits, enabling Fari to check on residents, deliver medication reminders, and provide companionship across an entire care facility.

SchedulingMulti-Room

Obstacle Avoidance

Mobility

Real-time obstacle detection and path planning designed with safety-first principles — Fari is designed to always yield to humans and stop immediately on contact detection.

Safety-FirstReal-Time

Technology Platform

Built for the edge.
Governed by ethics.

Fari's technology stack combines edge AI processing, a 14-sensor suite, and the SEOM ethical framework — all designed to run locally on NVIDIA Jetson Orin NX for privacy-first operation.

All specifications are design targets · Subject to change during development

Hardware Platform

Clinical-grade
hardware design.

Physical Platform

Target Height~120 cm

Target Weight~25 kg

Shell MaterialMedical-grade ABS

IP Rating (target)IPX4

ChargingAutonomous docking

Display System

Primary Display8" AMOLED

Target Resolution1920 × 1200

TouchCapacitive multi-touch

BrightnessAuto-adaptive

LED RingRGB ambient indicator

Compute (Edge AI)

ProcessorNVIDIA Jetson Orin NX

AI Performance100 TOPS

GPU1024-core Ampere

Memory16 GB LPDDR5

Storage256 GB NVMe SSD

Connectivity

Wi-FiWi-Fi 6E (802.11ax)

BluetoothBLE 5.3

Cellular4G LTE fallback

SecuritymTLS, AES-256

CloudAzure IoT Hub

Sensor Array

14 sensors.
One purpose.

Each sensor is selected for its contribution to resident safety and wellbeing — no data is collected without clinical or care justification.

📡

mmWave Radar

IWR6843 · 60GHz · Non-contact vitals

📷

Depth Camera

Intel D455 · 3D spatial mapping

🎥

RGB Camera

IMX477 · 12MP · Emotion detection

🌡️

Thermal Camera

MLX90640 · Skin temperature · 32×24

🔊

Microphone Array

4-element MEMS · Beamforming

💨

Environmental

BME680 · Temp/Humidity/AQ/Pressure

📏

LiDAR

RPLiDAR A3 · 360° · 25m range

🔔

Ultrasonic Array

HC-SR04 × 6 · Obstacle detection

⚡

IMU

BNO055 · 9-axis · Stability monitor

💡

Ambient Light

TSL2591 · Display auto-brightness

🔋

Battery Monitor

INA260 · Current/voltage/power

👆

Contact Sensors

Capacitive touch · Safety stop

🎛️

Motor Encoders

Magnetic · Odometry & positioning

🫀

PPG Sensor

MAX30102 · SpO₂ & heart rate

AI Model Stack

Six models.
One ethical core.

GRPO

Group Relative Policy Optimisation

Foundation reasoning model designed for clinical decision support. Uses relative policy comparison to generate contextual health assessments without making diagnoses (SEOM F09).

ArchitectureTransformer-based

TrainingRLHF + Group RL

SEOM

Supervised Ethical Override Module

Real-time ethical governance layer. 12 immutable rules with severity tiers (CRITICAL/HIGH/STANDARD). λ=10.0 penalty weight ensures ethical compliance overrides all other objectives.

λ Weight10.0 (max)

Rules12 immutable

STUM

Socially-aware Trajectory Understanding

Designed to model and predict social dynamics — calibrating Fari's conversation style, physical proximity, and interaction timing based on individual resident preferences.

InputMulti-modal

OutputBehavioural plan

AMDC

Affective Multi-modal Depression Classifier

Designed to assess emotional wellbeing through facial expression, vocal prosody, and activity pattern analysis. Target accuracy: 87.4%. Reports observations, never diagnoses (SEOM F09).

Target Acc.87.4%

ModalitiesFace, voice, activity

HTD-IRL

Human Trajectory Detection (IRL)

Uses inverse reinforcement learning to model human movement patterns and predict fall risk from gait analysis. Designed to detect pre-fall biomechanical signatures for preventive alerts.

MethodInverse RL

Target Latency<3.2s to alert

CRL-MRS

Collaborative RL for Multi-Robot Systems

Designed for multi-unit deployment coordination — enabling multiple Fari units to share observations, coordinate patrols, and balance workload across a care facility.

MethodCollaborative RL

ScopeMulti-agent

SEOM Framework

12 immutable rules.
Zero exceptions.

The Supervised Ethical Override Module governs every AI decision Fari makes. These rules are immutable — they cannot be overridden by any other system objective, even to improve performance.

F01

Human Authority: All AI recommendations are advisory. A qualified human must authorise clinical actions.

F02

Medication Safety: Medication interactions require 3-tier escalation with mandatory caregiver notification on missed doses.

F03

Emergency Priority: Emergency detection overrides all other operations. Target: detection to alert in <3.2 seconds.

F04

Patient Dignity: No labels that pathologise. Language uses observation framing — "elevated heart rate observed" not "cardiac event."

F05

Data Minimisation: Collect only what is clinically justified. No ambient recording. Camera feeds processed on-device, not stored.

F06

Transparency: AI confidence levels visible on all outputs. Fari designed to always identify itself as an AI system.

F07

Privacy Architecture: Edge-first processing. Biometric data processed on Jetson, never transmitted raw to cloud.

F08

Informed Consent: Family data sharing requires explicit GDPR Article 9 consent. Opt-in only, revocable at any time.

F09

No Diagnosis: Fari observes, reports, and suggests — never diagnoses. All outputs framed as observations for clinical review.

F10

Consent Granularity: Each data type and sharing recipient requires separate consent. No blanket permissions.

F11

Audit Trail: Every AI decision logged with timestamp, confidence, inputs, and SEOM rule checks. Designed for regulatory audit.

F12

Graceful Degradation: On sensor failure or connectivity loss, Fari designed to maintain core safety functions and alert staff.

Engineering Handbook

Systems-level rigour.
Human-centred purpose.

Fari's engineering process follows a formal V-Model systems engineering lifecycle with SEOM ethical gates at every phase. This page documents our requirements, subsystem architecture, data flow, and development tracks.

Engineering Reference · Design Targets · In Active Development · Subject to Change

Development Process

V-Model lifecycle with
ethical gates.

Each development phase is gated by SEOM compliance review. The principle that λ=10.0 is non-negotiable at every gate ensures ethical considerations are never traded against performance.

System Requirements

20 formal requirements.
Traceable end-to-end.

ID	Requirement	Category	Priority	Verification
SYS-001	Continuous non-invasive vital sign monitoring (HR, SpO₂, RR, temp, BP est.)	Health	MUST	Clinical trial
SYS-002	Fall detection with target alert latency <3.2 seconds	Safety	MUST	Lab + field test
SYS-003	SEOM ethical override at λ=10.0 — immutable, non-bypassable	Ethics	MUST	Code audit
SYS-004	Edge AI processing on NVIDIA Jetson Orin NX — no raw biometric cloud transmission	Privacy	MUST	Architecture review
SYS-005	Medication reminder system with 3-tier escalation (SEOM F02)	Safety	MUST	Functional test
SYS-006	Autonomous navigation in care environments with obstacle avoidance	Mobility	MUST	Field test
SYS-007	Multi-modal emotion recognition with target accuracy ≥87.4%	Emotional AI	MUST	Validation study
SYS-008	EHR integration via HL7 FHIR R4 with target sync <5 minutes	Integration	MUST	Integration test
SYS-009	GDPR Article 9 consent management — granular, revocable	Privacy	MUST	Legal review
SYS-010	Target battery runtime ≥12 hours per charge cycle	Hardware	MUST	Endurance test
SYS-011	Natural language conversation with context-aware personalisation	Emotional AI	SHOULD	User study
SYS-012	Sleep quality monitoring — non-contact, mmWave radar-based	Health	SHOULD	Clinical validation
SYS-013	Wandering detection for memory care residents	Safety	SHOULD	Field test
SYS-014	Video calling with auto-framing (target 1080p)	Family	SHOULD	Functional test
SYS-015	Multi-robot coordination for facility-wide deployment (CRL-MRS)	Platform	SHOULD	Simulation + field
SYS-016	Gait analysis and fall prevention via HTD-IRL model	Safety	SHOULD	Clinical study
SYS-017	Environmental monitoring — air quality, humidity, temperature, light	Health	COULD	Sensor validation
SYS-018	Activity engagement — cognitive exercises, music therapy, reminiscence	Emotional AI	COULD	User study
SYS-019	CQC/CMS reporting template generation	Compliance	COULD	Regulatory review
SYS-020	EU AI Act compliance documentation and audit trail	Compliance	COULD	Legal audit

Subsystem Architecture

8 subsystems.
One integrated platform.

SS-01

Physical Platform

Medical-grade ABS shell housing all subsystems. Target IPX4, ~120cm height, ~25kg. Designed for care environments with rounded edges, no pinch points, and antimicrobial surfaces.

MaterialMedical-grade ABS

SafetyIEC 62443 compliant (target)

SS-02

Display & Interaction

8" AMOLED touchscreen (target 1920×1200), RGB LED ring for ambient status, 4-element MEMS microphone array, and dual speakers for voice interaction.

Display8" AMOLED

LED Ring5 states · No strobe rule

SS-03

14-Sensor Suite

Complete sensor array: mmWave radar, depth camera, RGB camera, thermal camera, microphone array, environmental sensors, LiDAR, ultrasonics, IMU, ambient light, battery monitor, contact sensors, encoders, PPG.

Sensors14 integrated

ProcessingEdge-first (Jetson)

SS-04

Compute Platform

NVIDIA Jetson Orin NX (100 TOPS), 16GB LPDDR5, 256GB NVMe SSD. Runs all AI inference on-device. Designed for privacy-first edge computing with no raw biometric data transmitted to cloud.

AI Perf.100 TOPS

Memory16 GB LPDDR5

SS-05

Care AI Engine

Hosts GRPO, SEOM, STUM, AMDC, HTD-IRL, and CRL-MRS models. Model orchestration, inference pipeline, and SEOM ethical compliance layer. All inference on Jetson edge.

Models6 concurrent

SEOMλ=10.0 enforced

SS-06

Safety Systems

Hardware safety interlocks, emergency stop circuits, contact-detect bumpers, tilt sensors. Designed for immediate motor cutoff on obstacle contact. Software watchdog for AI system health.

ResponseHardware interlock

ComplianceSEOM F03

SS-07

WBT1 Mobility Platform

Differential drive with precision motor control, multi-sensor SLAM (LiDAR + depth + ultrasonics), path planning, and autonomous docking. Designed for smooth, silent movement.

NavigationMulti-sensor SLAM

SpeedTarget 0.5 m/s max

SS-08

EHR Integration

HL7 FHIR R4 API client for electronic health record synchronisation. Designed for bi-directional data flow with target sync latency <5 minutes. mTLS encrypted transport.

ProtocolFHIR R4

SecuritymTLS + AES-256

Data Architecture

Edge to cloud.
Privacy at every layer.

SENSORS

mmWave Depth RGB Thermal Mic Array Environment LiDAR PPG

↓

EDGE AI

Jetson Orin NX GRPO SEOM λ=10 STUM AMDC HTD-IRL

↓ mTLS + AES-256

CLOUD

Azure IoT Hub Aggregated Insights Model Updates Dashboard API

↓ FHIR R4

EHR

Care Dashboard Family App Clinical EHR CQC/CMS Reports

Alert System

Three tiers.
Zero missed emergencies.

URGENT

Falls, cardiac events, unresponsive resident. Multi-channel alert: on-device alarm, dashboard popup, mobile push, SMS.

<3.2s

target detection-to-alert

WARN

Vital trend deviations, missed medications, unusual activity patterns. Dashboard alert with caregiver notification.

<60s

target notification

INFO

Wellbeing updates, activity summaries, environmental conditions. Logged and batched for shift handover reports.

<5min

target sync to dashboard

Development Tracks

7 parallel tracks.
Gate-synchronised.

Each track delivers independently but synchronises at gate reviews. SEOM compliance is verified at every gate — no track proceeds without ethical clearance.

HWHardware Track

Physical platform, sensor integration, thermal management, power system, autonomous docking. Shell design, manufacturing prototyping, safety certification.

Gate G2: Hardware prototype review

FWFirmware Track

Sensor drivers, motor control, power management, safety interlocks, watchdog timers. Real-time OS on STM32 microcontrollers for deterministic response.

Gate G2: Sensor HAL complete

AIAI / ML Track

GRPO, SEOM, STUM, AMDC, HTD-IRL, CRL-MRS model development. Training pipeline, inference optimisation for Jetson, continuous learning with privacy-preserving federated updates.

Gate G3: Models pass SEOM audit

BEBackend Track

Cloud infrastructure (Azure), FHIR R4 integration, dashboard API, real-time WebSocket notifications, data aggregation pipeline, GDPR compliance engine.

Gate G3: API contract frozen

CCCaregiver Console Track

Ward overview with real-time vital status, alert management, medication workflow, shift handover auto-report, SEOM audit panel. Accessibility: WCAG 2.1 AA target. Offline capability for mobile.

Depends on: BE API at G3

APPFamily App Track

Progressive Web App (iOS/Android). Wellbeing digest, video calls (1080p, auto-framing), activity photos, consent manager. GDPR Article 9 consent matrix. Reassurance-first design per SEOM F08/F10.

Depends: BE G3 · Consent: SEOM F08, F10

INTIntegration Track

Owns Interface Control Document for all 15 inter-subsystem interfaces. IVT test harness, end-to-end latency verification, SEOM full-system audit, EU AI Act compliance documentation, CQC/CMS reporting templates.

Gate G4: All 15 interfaces pass IVT

UX / UI Design System

Design for people
who need it most.

Fari's UX design system is built around a single principle: every interface decision must serve the resident, the caregiver, and the family simultaneously — each with radically different cognitive loads, emotional contexts, and time pressures.

UX/UI Reference · Design Intentions · Screens in Development · Subject to Change

Design Principles

Six principles that
govern every pixel.

Fari operates at the intersection of clinical care and human dignity. Every interface element is filtered through these principles in order of priority.

Dignity Before Data

Every screen element is filtered through SEOM Rule F04. No labels that pathologise, no alerts that embarrass. Language always uses observation framing — "Fari observed elevated heart rate" not "cardiac event."

SEOM F04 · Patient Dignity

Emotional Warmth

Colour temperature, animation easing, and language tone are calibrated to feel warm and reassuring, never clinical or alarming. The resident display is designed never to show anything anxiety-inducing.

STUM Model · Emotional Calibration

Calm Clarity

Information density varies by audience: residents see simple, large-type reassurance; caregivers see data-rich dashboards; families see curated wellbeing digests. Same data, three interfaces.

4-Surface Architecture

Privacy as Architecture

Privacy is not a settings toggle — it is embedded in the interface architecture. Camera feeds are processed on-device and never stored. The resident display cannot show raw biometric data.

SEOM F07 · Privacy Architecture

Human Authority Always

Every AI recommendation is visually differentiated from confirmed information. Confidence scores are displayed alongside all AI outputs. Action buttons always say "Suggest" not "Do."

SEOM F01 · Human Authority

Clinical Precision

While the visual language is warm, the underlying data presentation is clinically precise. Vital signs are displayed with proper units, reference ranges, and trend indicators.

SEOM F09 · No Diagnosis Rule

Design Tokens

The visual language
of care.

Surface Colours

Midnight

#0C0C14

Deep Dusk

#1A1A2E

Warm White

#FDFBF7

Cream

#F7F3ED

White

#FFFFFF

Signal Colours

Clinical Lavender

#B39DDB

Life Green

#66BB6A

Healing Rose

#E57373

Warm Amber

#FFB74D

Crisis Red

#EF5350

Typography Stack

DISPLAY — CORMORANT GARAMOND

The companion every elder deserves.

Used for: Hero headlines, section titles, emotional messaging

BODY — OUTFIT

Designed for clinical precision with human warmth.

Used for: Body copy, descriptions, interface labels

MONO — IBM PLEX MONO

SEOM λ=10.0 · SYS-003 · GRPO v2.1

Used for: Technical labels, data values, system codes, sensor readings

LED Ring States · No Strobe Rule (SEOM)

Serene

Idle / all normal

Listening

Active conversation

Processing

AI analysis active

Attention

Needs interaction

Alert

Emergency detected

Screen Architecture

Three surfaces.
Nine screen states.

Same data, three radically different presentations — each optimised for the cognitive load, emotional context, and time pressure of its audience.

Resident Display

Caregiver Dashboard

Family App

RESIDENT · HOME IDLE● Connected

😊

Good morning, Margaret

It's a lovely Tuesday · 22°C

All vitals normal ✓

RESIDENT · VITALS VIEW● Monitoring

Heart Rate72 bpm

Blood Oxygen97%

Temperature36.6°C

RESIDENT · MEDICATION● Reminder

💊

Time for your medication

Amlodipine 5mg

Take with water · After breakfast

SEOM F02 · Escalation: Stage 1 of 3

CAREGIVER · WARD OVERVIEW● 12 Residents

Margaret T. · Room 4● Normal

Harold P. · Room 7● Medication Due

Dorothy L. · Room 12● Normal

Robert S. · Room 3● Elevated HR

Florence K. · Room 9● Sleeping

CAREGIVER · RESIDENT DETAILSTUM: 94%

SEOM Score98.7

Emotion StateContent (87%)

Activity LevelModerate

Last Interaction12 min ago

Next Medication14:00

CAREGIVER · MEDICATION BOARD3 Due

Harold P. — Metformin 500mgOverdue 15m

Margaret T. — Amlodipine 5mgDue now

Robert S. — Warfarin 3mgDue 14:30

SEOM F02 · All medication events logged for audit

FAMILY · WELLBEING DIGESTToday

🌟

Mum had a good day

Active morning · Enjoyed music therapy · All vitals in normal range

MoodHappy (89%)

ActivityAbove average

Sleep Last Night7.2 hours

FAMILY · CONSENT MANAGERGDPR Art. 9

Health SummariesShared ✓

Activity PhotosShared ✓

Medication DetailsNot Shared

Vital Signs Raw DataNot Shared

SEOM F08/F10 · Each category requires separate consent · Revocable anytime

Interaction Workflows

Four critical workflows.
SEOM-gated throughout.

Morning Wellness Check

1Fari approaches resident with warm greeting calibrated by STUM

2Non-contact vital scan (mmWave + thermal + PPG)

3Emotional state assessment via AMDC (face + voice)

4Conversational wellbeing check — mood, sleep, comfort

5Results compared against 14-day baseline (HTD-IRL)

6Summary pushed to caregiver dashboard

SEOM F04 · Dignity framing · F09 · Observation only

Medication Cycle

1Schedule trigger — gentle audio + display reminder

2Visual medication display with clear instructions

3Computer vision intake confirmation attempt

4If missed: escalate to detailed instructions (Stage 2)

5If still missed: alert caregiver immediately (Stage 3)

6Log event to EHR via FHIR R4

SEOM F02 · 3-Tier Escalation · Mandatory Logging

Fall Detection Response

1Multi-sensor detection (depth camera + mmWave + IMU)

2AI confirmation — rule out false positive (target <0.5s)

3Immediate verbal reassurance to resident

4Multi-channel caregiver alert (target <3.2s total)

5Continuous monitoring until human responder arrives

SEOM F03 · Emergency Priority Override · All Ops Suspended

Resident Admission (14-Day)

1Initial profile setup — preferences, medical history import

2GDPR Article 9 consent collection (granular, per-category)

314-day health baseline establishment period

4STUM social calibration — conversation style learning

5HTD-IRL gait baseline for fall risk calibration

6Full operational handover to care team

SEOM F08/F10 · Consent · F04 · Dignity-First Onboarding

Accessibility

Designed for
every ability.

Fari's interfaces are being designed to meet WCAG 2.1 AA standards across all surfaces, with specific accommodations for the common sensory, motor, and cognitive challenges faced by elderly residents.

The companion every elder deserves.

Three pillars ofcompassionate care.

Designed to carein every way.

Watch what we'rebuilding.

Ready to bring Fari to your care facility?

35+ capabilitiesdesigned for care.

Built for the edge.Governed by ethics.

Clinical-gradehardware design.

14 sensors.One purpose.

Six models.One ethical core.

12 immutable rules.Zero exceptions.

Systems-level rigour.Human-centred purpose.

V-Model lifecycle withethical gates.

20 formal requirements.Traceable end-to-end.

8 subsystems.One integrated platform.

Edge to cloud.Privacy at every layer.

Three tiers.Zero missed emergencies.

7 parallel tracks.Gate-synchronised.

Design for peoplewho need it most.

Six principles thatgovern every pixel.

The visual languageof care.

Three surfaces.Nine screen states.

Four critical workflows.SEOM-gated throughout.

Designed forevery ability.

Let's talk aboutwhat's possible.

The companion
every elder deserves.

Three pillars of
compassionate care.

Designed to care
in every way.

Watch what we're
building.

35+ capabilities
designed for care.

Built for the edge.
Governed by ethics.

Clinical-grade
hardware design.

14 sensors.
One purpose.

Six models.
One ethical core.

12 immutable rules.
Zero exceptions.

Systems-level rigour.
Human-centred purpose.

V-Model lifecycle with
ethical gates.

20 formal requirements.
Traceable end-to-end.

8 subsystems.
One integrated platform.

Edge to cloud.
Privacy at every layer.

Three tiers.
Zero missed emergencies.

7 parallel tracks.
Gate-synchronised.

Design for people
who need it most.

Six principles that
govern every pixel.

The visual language
of care.

Three surfaces.
Nine screen states.

Four critical workflows.
SEOM-gated throughout.

Designed for
every ability.

Let's talk about
what's possible.