Machine learning platform to scale pet health monitoring
.png)
<2 mo
To high-accuracy image analysis
10X
Human speed with image analysis
Dozens
Of machine learning models deployed
TL;DR
.svg)
DIG Labs needed a machine learning platform that could analyze pet health from photos across many pet types and indicators, handle tens of thousands of daily requests, and support custom workflows per enterprise client.
We built three products: a SageMaker ML pipeline with expert review, a multi-tenant platform with automated onboarding, and an API-first product for pet, vet and health companies.
High-accuracy image analysis arrived in under two months, at 10× human speed, with dozens of models deployed. DIG Labs was later acquired.
Take a picture of your pet, and instantly know what to do. That is the DIG Labs promise. We built a visual AI engine, trained on hundreds of thousands of images, that analyzes pet health across dozens of areas such as ears, nose and teeth.
Those insights then drive personalized plans such as diet transitions and weight loss programs. DIG Labs was later acquired.
The challenges
The platform had to be both a serious ML system and an enterprise product.
1. Complex proprietary datasets
2. Scale
3. Multi-tenant customization
Each enterprise client might need custom workflows, or even its own proprietary model.
What we built
1. The machine learning pipeline
The ML pipeline runs as its own application, separate from the platform, so teams move faster and isolate problems. Athena queries large datasets, PostgreSQL manages metadata and S3 stores images. SageMaker Pipelines orchestrate the lifecycle with continuous training, Processing Jobs prepare data and Training Jobs handle training and hyperparameter optimization.
Every model is registered and A/B tested, and domain experts review each one against model cards before approval. Approved models deploy through a RESTful API for real-time insights from iOS and web. The modular design cut the time to scale tenant-specific models by up to five times.
.png)
2. The multi-tenant platform
The platform consumes the ML APIs at scale, with each tenant getting its own data requirements, model configuration and workflows behind an API Gateway. A turnkey white-label app deploys in seconds, and tenant onboarding through a Retool app dropped from over 12 hours to under 10 minutes. Integrations include Shepherd, Segment, Airtable and Gladly.
3. An API-first product
Enterprises across pet, vet and health built custom experiences on the RESTful API, documented with Swagger and Postman from YAML files, with AWS API Gateway managing credentials and rate limits.
Conclusion
The platform gave DIG Labs and its clients highly accurate pet image analysis through a modular ML pipeline and strong feedback loops. Automated onboarding, white-label and custom configurations, and an open API-first architecture supported fast iteration and scalable adoption.
Building an ML platform that has to serve many clients at once? Get in touch.
<2 mo
To high-accuracy image analysis
10X
Human speed with image analysis
Dozens
Of machine learning models deployed
Use case: visual AI, engineering, design | Industry: pets
ML: SageMaker Pipelines and Studio, model registry, A/B testing, expert review with model cards
Data: AWS Athena, PostgreSQL, Amazon S3 | Serving: RESTful API via AWS API Gateway
Platform: multi-tenant, white-label app, Retool onboarding | Docs: Swagger, Postman
FAQ
How do you build a multi-tenant machine learning platform?
Separate the ML pipeline from the tenant platform, serve models through a RESTful API behind an API Gateway, and give each tenant its own data requirements, model configuration and workflows.
How does visual AI for pet care stay accurate in production?
Register every model, A/B test it, and have domain experts review it against model cards before deployment, with continuous training through SageMaker Pipelines.
How fast can a new enterprise tenant be onboarded?
Automated onboarding through a Retool app cut setup from over 12 hours to under 10 minutes.
What results did the platform deliver?
High-accuracy image analysis in under two months, at 10× human speed, with dozens of models deployed. See our AI agents work.
Athena, PostgreSQL, S3
API Gateway, Retool
Read more
You’re all set!
.svg)
