Success Story
Emi Labs modernizes its platform, consolidates 98% of redundant network resources and optimizes its cloud costs
At a glance
Emi Labs is an HR Tech company with recruiting platforms that handle high volumes of traffic and data. Its infrastructure was rigid and costly, with mixed environments, redundant network resources and fragmented databases. Over three projects, Craftech modernized the platform on Kubernetes, migrated the databases to a highly available managed engine and optimized cloud costs: redundant network resources were cut by 98%.
98%
fewer redundant network resources
80+ → 1
network resources consolidated per environment
USD 100+/mo
savings per optimized server
Emi Labs is an HR Tech company with recruiting platforms: multiple applications for tracking and hiring staff, with high volumes of traffic and data.
Solution implemented
- Modernized the platform toward a scalable architecture on Kubernetes
- Set up isolated development, testing and production environments
- Automated the release of new versions in a fast, secure and standardized way
- Implemented capacity that adjusts automatically to demand
- Migrated the databases to a managed cloud engine, with tailored strategies to avoid prolonged outages
- Ran a FinOps audit, optimized a flow of unnecessary downloads and applied rightsizing and Spot Instances
Business Challenge
Emi Labs' infrastructure was rigid and costly. It mixed test and production environments, putting real data at risk, and dozens of redundant load balancers inflated the bill. On top of that came unexplained cost spikes, mainly from network data transfer, servers left running and oversized, and databases fragmented across multiple independent servers with no high-availability guarantees. The team needed to ship features without fear of breaking the platform.
Solution
Craftech worked with Emi Labs on three projects. First, it modernized the platform toward a scalable architecture on Kubernetes, with isolated environments and automated deployments. Then, it migrated the databases to a highly available managed cloud engine. Finally, it ran a FinOps audit that eliminated the data transfer overspend and matched capacity to actual consumption.
Outcomes
- 98% fewer redundant network resources: more than 80 consolidated into one per environment
- Developers who test without touching real data
- Much shorter deployment times
- Databases with high availability and fault tolerance, and more than USD 100/month in savings per optimized server
- The data transfer overspend eliminated
- Automatic node consolidation every 5 minutes to shut down empty servers, and cost dashboards per application
Modernization and scalability (December 2023 – June 2026)
The infrastructure mixed test and production environments, putting real data at risk, and had dozens of redundant load balancers that inflated the bill. Craftech fully modernized it toward a scalable architecture on Kubernetes.
- Isolated development, testing and production environments
- Automation to release new versions in a fast, secure and standardized way
- Capacity that adjusts automatically to demand
- 98% fewer redundant network resources: more than 80 consolidated into one per environment
- Developers who test without touching real data, and much shorter deployment times
High-availability databases (September 2024 – February 2025)
The databases were fragmented across multiple independent servers: they were complex, costly and had no high-availability guarantees. Craftech migrated them to a managed cloud database engine, with tailored strategies to avoid prolonged outages.
- High availability and fault tolerance
- More than USD 100/month in savings per optimized server
- A transition that was transparent to end users
FinOps (August – September 2025)
The bill showed unexplained spikes, mainly from network data transfer, and servers were left running and oversized. Craftech ran a FinOps audit, found processes that downloaded unnecessary files from the internet thousands of times a day, and optimized that flow.
- Rightsizing based on historical consumption and automatic use of Spot Instances
- The data transfer overspend eliminated
- Automatic node consolidation every 5 minutes to shut down empty servers
- Cost dashboards per application
Stack & technologies
