AviIQ Platform Architecture

IoT + AI monitoring stack for poultry farms

Temp / Humidity
Ammonia
CO2
Cameras
Thermal
MQTT Gateway
MQTT + Stream Processing
FlockMonitoringPipeline
TopicsConsumersStream ProcessorsSchedulesTelemetry
Consumer Queues
AviIQ Cluster
Telemetry Ingestion
Sensor time-series
tq:telemetry-ingestion
Environmental Analysis
Anomaly detection
tq:environmental-analysis
Vision / Image Analysis
Thermal CV / CNN
tq:vision-analysis
Alert Orchestrator
Rules / Notifications
tq:alert-orchestrator
TimescaleDB (telemetry)
Object Storage (images)
Model Registry
PostgreSQL

Core components

Six layers that turn farm sensor and camera data into actionable flock-health insight.

01
Data ingestion layer

Sensors, cameras, and thermal via on-farm MQTT gateway

AviIQ ingests farm sensor and camera streams and turns them into analysis-ready telemetry and image frames.

Sensors & cameras
SHT31 Temp/HumidityMH-Z19C CO2MQ-137 AmmoniaSPS30 DustBH1750 LightONVIF CameraRTSP StreamThermal CameraWater MeterFeed SensorWeigh ScaleDoor / MotionFan / ActuatorMQTT Gateway
Event triggers
new_telemetry_batchimage_frame_capturedammonia_threshold_exceededthermal_scan_completeweight_sample_ready
Ingestion engine
  • Sensors + ONVIF/RTSP cameras stream via on-farm MQTT gateway
  • TimescaleDB for telemetry, object storage for image frames
  • Topic contracts and schema registry per device type
  • Quality checks on every telemetry batch before processing
02
Feature engineering

Time-series and image-frame feature pipelines

AviIQ builds rich feature sets from sensor time-series and camera frames across every shed and flock.

  • Environmental features from SHT31 temp/humidity, MH-Z19C CO2, and MQ-137 ammonia trends
  • Air-quality features from SPS30 dust and BH1750 light readings
  • Image frames extracted and normalized from ONVIF/RTSP and thermal cameras
  • Flock-level aggregates rolled up from device telemetry per shed and house
03
AI model services

Fever detection, feces CNN, weight and movement CV

AviIQ deploys the right model for each signal—from thermal computer vision to environmental anomaly detection.

  • Fever detection from thermal camera imagery using thermal computer vision
  • Feces classification CNN surfacing early disease indicators
  • Weight estimation and movement tracking from camera frames
  • Environmental anomaly detection over sensor time-series
04
Orchestration engine

Event pipeline and alert rules engine

AviIQ routes events from MQTT through stream processing to model services, then applies alert rules that decide what needs attention.

  • Durable event pipeline: MQTT → stream processing → model services
  • Alert rules engine over telemetry thresholds and model outputs
  • Coordination across ingestion, environmental, and vision services
  • Prioritization of flock health, environment, and device-fault events
05
Dashboard & portal

Web dashboards and multilingual farmer mobile app

AviIQ gives managers and farmers a live view of every shed, flock, and device—on web and mobile.

  • Web dashboards across organization, farm, shed, flock, and device views
  • Live sensor charts, camera feeds, and thermal overlays per shed
  • Multilingual farmer mobile app with real-time push alerts
  • Alerts delivered via SMS (MSG91), email, and push (FCM)
06
Human-in-the-loop gates

Vet and manager approval with full audit trail

Every high-impact action goes through human review—and every decision is logged for audit.

  • Vet or manager approval for isolation, medication, and culling recommendations
  • Recommendations recomputed instantly as flock and shed data changes
  • Action handoff to farm operations with validation before dispatch
  • Full audit trail from AI detection through vet override to action taken

Flock Health Approval Gate

Vet or farm manager must review high-impact actions
in flock health monitoring workflows

ShedIsolationActionWorkflow
Running
HITL Gate

Health Action Review

Waiting for Vet Approval
DR
Farm VeterinarianHealth Action Approval
Timer

"Isolate Shed 3"
queued…

Activity

Dispatch Shed Isolation

DetectionImpact CheckVet ReviewAction Dispatched
9:01 AMHealth anomaly detectedcompleted2m9:03 AM
9:03 AMVet review requestedcompleted5m9:08 AM

Every AI decision can be reviewed — and replayed.

AviIQ persists complete monitoring histories. Every AI detection, alert, vet or manager action, and health outcome is logged in the decision trail.

  • Full decision provenance — which module flagged what, when, and from which sensor or camera
  • Vet & manager feedback recorded — accepted, modified, or rejected with reasoning
  • Outcomes tracked — compare predicted vs actual flock health impact of every action

Health Decision Log + Replay

Audit flock health decisions and replay monitoring history

FlockHealthWorkflow
Completed
Flock IDVetFarm Veterinarian
StatusCompleted1h– 56 %
Activity Log
Telemetry Ingestioncompleted 4m9:01 AM
Environmental Analysiscompleted 8m9:05 AM
Thermal Scanescalated in 15m9:13 AM
Vet Reviewcompleted 12m9:28 AM
Alert Dispatchcompleted 3m9:40 AM
DR
Health Action ReviewFarm Veterinarian
Health Impact+8.3%
Mortality −3.1%flock mortality
FCR −4.2%feed conversion ratio
Vet DecisionApproved · Modified
DR
Health Actions
  • Shed 3: ammonia 28ppm → increase ventilation
  • Flock B: 4 birds flagged for inspection
  • Shed 3: thermal scan → 2 birds with fever
  • Vet: Approved with modification
8:39 AM8:50 AM
Telemetry Ingestion
Environmental Analysis
Thermal Scan
Vet Review
Alert Dispatch

Ready to bring intelligence to your flock?

AviIQ is not a tool you configure once. It's a living intelligence layer for your farm that gets smarter every day.