Workflow · Production

Cubo Ai ensures infant safety at scale with Google Cloud AI infrastructure

The problem

Cubo Ai's multicloud environment became increasingly costly as user growth accelerated in 2019, and its first-generation baby monitor lacked real-time device data, making remote troubleshooting impossible.

First attempt

The multicloud setup led to rising costs that were unsustainable as user volumes grew, causing Cubo Ai to abandon it in favour of a single cloud platform.

Workflow diagram · grounded in source
1
AI camera detects sleep risk
ai_action
“can detect a covered face, if babies are sleeping on their stomach or on the side, as well as other dangers in babies' sleep and living environments with an accuracy rate of up to 92%”
2
Alert sent through Firebase
integration
“an alert is sent from the AI camera to the application through Firebase”
3
Load balancing routes to nearest data center
routing
“It uses Cloud Load Balancing to connect users to the nearest data center when an alert is sent from the AI camera to the application”
4
Parent receives app alert
output
“parents receive an alert on an app, so they can attend to the baby immediately”
5
IoT Core streams device status
integration
“uses IoT Core to send device status data in real time. The information is then analyzed on Cloud Logging for Cubo Ai's engineers to clear log errors remotely”
6
BigQuery generates feature reports
integration
“the data sent back from devices are analyzed on BigQuery to generate detailed reports illustrating the performance of the new feature. Cubo Ai's research and development team can then quickly adjust the new feature based on the reports, …”
Reported outcome

After consolidating on Google Cloud, Cubo Ai achieved zero cloud-related downtime over two years, handled more than 10X user growth with the same IT workforce, reduced container unit costs by 20%, and shortened feature development time from several weeks to less than one week.

Reported metrics
AI detection accuracy rateup to 92%
Daily active usersaround 70,000
user growth managed with same IT workforcemore than 10X user growth
Container unit cost reduction20%
Show all 8 reported metrics
AI detection accuracy rateup to 92%
daily active usersaround 70,000
user growth managed with same IT workforcemore than 10X user growth
container unit cost reduction20%
operational cost reduction from GPU optimization5%
cloud-related downtimeZero downtime in two years of deployment
feature development timefrom several weeks to less than one week
alert response time reductionat least one second
Reported stack
Google CloudIoT CoreCloud FunctionsCloud Load BalancingCloud GPUsGoogle Kubernetes EngineCloud LoggingBigQueryFirebaseCloud StorageCloud SQLKubernetesCloudMile
Source
https://cloud.google.com/customers/cubo-ai
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

After consolidating on Google Cloud, Cubo Ai achieved zero cloud-related downtime over two years, handled more than 10X user growth with the same IT workforce, reduced container unit costs by 20%, and shortened featur…

What tools did this team use?

Google Cloud, IoT Core, Cloud Functions, Cloud Load Balancing, Cloud GPUs, Google Kubernetes Engine, Cloud Logging, BigQuery, Firebase, Cloud Storage.

What results were reported?

AI detection accuracy rate: up to 92%; Daily active users: around 70,000; user growth managed with same IT workforce: more than 10X user growth; Container unit cost reduction: 20% (source-reported, not independently verified).

What failed first in this deployment?

The multicloud setup led to rising costs that were unsustainable as user volumes grew, causing Cubo Ai to abandon it in favour of a single cloud platform.

How is this workflow AI workflow structured?

AI camera detects sleep risk → Alert sent through Firebase → Load balancing routes to nearest data center → Parent receives app alert → IoT Core streams device status → BigQuery generates feature reports.