Client: Freepik Company
Sector: Technology / Creative Content
Freepik Company, one of the largest graphic resource platforms in the world, faced a key challenge: overcoming the limitations of a content ingestion system that hindered the company's growth.
Challenge
Freepik Company faced a critical bottleneck in its content ingestion system. Its monolithic infrastructure based on virtual machines could barely process 2 million files per month, with a 20% error rate in the received resources and a huge manual workload to requeue, correct, and validate the content. The situation not only limited growth but also generated high operational costs and a significant accumulation of technical debt.
Solution
A completely new architecture based on microservices deployed on Google Kubernetes Engine (GKE) was designed and led the implementation. This new solution was capable of receiving and processing up to 40 million files per month —including PSDs, EPS, JPGs, among others— applying a series of automatic validations:
NSFW content detection
Automated quality filters
Preview generation
Metadata and consistency validations
User correction history
Additionally, a human-in-the-loop strategy was implemented for cases that required specialized review. This interaction was managed through a React webapp, designed to provide agile feedback to specific users within the workflow.
Implementation
The microservices, developed in Python, were ready to scale completely autonomously and horizontally, dynamically adapting to the load volume. The databases managed hundreds of millions of records with efficiency and resilience.
The system not only drastically increased ingestion capacity but also reduced processing times from hours to seconds per resource, eliminating bottlenecks and significantly decreasing manual errors.
Impact
From 2M to 40M files processed monthly
Reduction of errors from 20% to marginal figures
Automation of critical processes that previously required manual intervention
Significant savings in infrastructure and support costs
Solid foundation to continue scaling the resource catalog
