Joot built a scalable visual AI platform on AWS using the Serverless Framework, achieving 70-90% server cost reduction and completing their full product in just 6 months.
Massive reduction in server costs by eliminating idle capacity and paying only for actual usage with a serverless architecture.
Completed their full visual AI software platform in just 6 months using the Serverless Framework and AWS services.
Avoided resource-limited machine challenges and saved time on DevOps provisioning, letting the team focus entirely on product.
Reduced coding effort through serverless composition, connecting managed AWS services instead of writing custom infrastructure code.
Joot helps content creators and advertisers predict and improve social media and advertising image engagement using machine learning and AI. With a small but highly skilled engineering team, Joot needed an architecture that would allow them to develop fast, get something up and running quickly, and design their system with existing managed services that would just work.
Joot analyzes millions of data points to rank and score images, providing a prediction tool that helps businesses achieve higher ROI, conversions, and engagement on social media and advertising.
By adopting a fully serverless architecture on AWS with the Serverless Framework from the very beginning, Joot was able to construct everything they needed within a 6 month period, focusing their engineering effort on building product features rather than managing infrastructure.
“We didn't start off Joot by trying something else and then deciding to go with Serverless. This was a conscious decision from the very beginning. We needed to develop fast, get something up and running and to be able to design our architecture with existing services that would just work.”
Chris Crabtree
Co-founder, Joot
As a growing startup building visual AI software, Joot faced several critical challenges:
The team did not want to depend on DevOps or worry about whether they had enough resources available. Managing servers would slow down development.
Traditional infrastructure meant dealing with resource-limited machine challenges, requiring constant provisioning and capacity planning.
Joot needed to build a complete visual AI platform from scratch and get it to market as quickly as possible with a lean team.
The platform needed per-customer machine learning models for image scoring, requiring a flexible architecture that could handle diverse compute needs across Node.js and Python workloads.
Joot adopted a fully serverless architecture on AWS, using the Serverless Framework to define, deploy, and manage their entire infrastructure as code. This was a conscious decision from day one, not a migration from another approach.
Users sign up via Cognito and interact with an Amplify-powered web application. When users upload images, they are stored in S3, which triggers Lambda functions for image resizing and metadata extraction. DynamoDB stores the image metadata, while EventBridge routes events to SQS queues that are processed by Lambda functions. SageMaker provides per-customer ML model endpoints for image scoring, delivering personalized predictions for each user.
The team uses Node.js for general application development and Python for machine learning workloads, taking advantage of Lambda's multi-runtime support.
Define all infrastructure in serverless.yml files, enabling version control, code review, and repeatable deployments.
Deploy isolated environments for development, staging, and production with a single command.
Leverage event-driven patterns with SQS, EventBridge, and S3 triggers for asynchronous image processing.
Integrate per-customer machine learning models via SageMaker endpoints called from Lambda functions.
Reduce coding effort through serverless composition, connecting managed services instead of writing custom infrastructure code.
Support multi-language workloads with Node.js for general development and Python for ML.
“One of the number one factors we chose Serverless was because we didn't want to depend on dev-ops and worry about whether we had enough resources available.”
Chris Crabtree
Co-founder, Joot
Authentication & user management
REST API management & routing
Serverless compute for business logic
Frontend web application hosting
Image storage triggering Lambda functions
Message queuing for async processing
Event routing to SQS queues
Per-customer ML model endpoints for image scoring
Metadata storage for images and application data
Languages: Node.js for general application development, Python for ML workloads.
By eliminating always-on servers and paying only for actual compute usage, Joot achieved a 70-90% reduction in server costs compared to traditional infrastructure. The pay-per-use model aligns perfectly with their variable workload patterns.
Joot was able to construct everything they needed within a 6 month period, building their entire visual AI software platform from scratch using the Serverless Framework and AWS managed services.
The team avoided the overhead of DevOps provisioning entirely. No servers to patch, no operating systems to maintain, and no scaling policies to tune. Engineers dedicate their time to building product features that drive business value.
By composing managed AWS services together through the Serverless Framework, Joot reduced the amount of custom code they needed to write. S3 triggers, EventBridge routing, and SQS processing replaced what would have been significant custom infrastructure code.
The serverless architecture scales automatically with demand. During traffic spikes, Lambda provisions additional capacity instantly, and scales back to zero during quiet periods. No capacity planning, no resource-limited machine challenges.
“We were able to construct everything we needed within a 6 month period of time and it was wonderful. Kudos to the Serverless team, our experience at Joot was just phenomenal being able to use Serverless to build this whole application from scratch.”
Chris Crabtree
Co-founder, Joot
Building on their success with the Serverless Framework, Joot has expanded to offer finely-tuned AI models across 100+ industries. The serverless architecture that enabled them to build their initial platform in 6 months continues to support their growth, scaling effortlessly as they onboard new customers and industries without any changes to their underlying infrastructure.
Join Joot and thousands of other teams building on the Serverless Framework. Deploy your first serverless application in minutes.