Automating Kubernetes Deployments for an AI Security Platform
DBB Software codified Lakera's cloud infrastructure and built a GitOps deployment pipeline on Kubernetes, so the team ships new services to production safely and on demand across two clouds.
Industry
Technology
Service
Infrastructure Services
Team
1 DevOps Engineer
Project State
June 2024 - Ongoing
Country
United States
About the Client
Lakera is an AI-native security platform that protects generative AI applications and agents from threats such as prompt injection, data leakage, and model abuse. Its products give enterprises runtime protection, AI red teaming, and workforce AI security, so teams can deploy GenAI without opening up new attack surfaces.
The Client's Initial Request
Lakera engaged DBB Software to automate its cloud infrastructure and deployments, so a fast-growing AI security platform could ship more often without trading away reliability or cost control.
Faster, Safer Deployments
Replace manual release steps with a pipeline that ships changes to production quickly and predictably.
01
Scalability on Demand
Let the platform absorb growing load without re-provisioning infrastructure by hand each time.
02
Cost Control
Keep cloud spend in check as the platform and its workloads grow.
03
Production Reliability
Keep a high-traffic security service stable while new services roll out.
04
Solutions We Delivered
DBB Software embedded a DevOps engineer inside Lakera's team to codify the platform's infrastructure and automate how services reach production.
Kubernetes Platform on Amazon EKS
The team configured and maintains Lakera's Amazon EKS clusters and packages each service as a Helm chart for repeatable deployment. New services land on the cluster the same way every time, which keeps the platform predictable as Lakera adds capability.
End-to-End Delivery Pipeline with GitLab and ArgoCD
A GitLab CI pipeline builds and tests each change, packages it as a Docker image, and pushes it to the registry, then ArgoCD deploys it to the cluster. Releases flow through Argo Rollouts for controlled, reversible rollouts, so a bad change can be backed out without a scramble.
Infrastructure as Code Across AWS and GCP
All infrastructure is provisioned and managed through Terraform across AWS and GCP, with AWS Secrets Manager holding credentials and ECR and GCR serving as image registries. Every change to the environment is reviewed and version-controlled instead of clicked together in a console, so the setup stays auditable and reproducible.
Observability and Cost-Aware Autoscaling
The platform is integrated with Grafana, Prometheus, Loki, and Sentry for metrics, logs, and error tracking, and Karpenter scales cluster capacity to match actual demand. The team can see how the system behaves and Lakera pays for the compute it actually uses.
Results Achieved
Bad Changes Back Out Cleanly
Controlled rollouts let a problem release be reversed quickly, so a single bad deploy doesn't turn into an outage.
Capacity Follows Demand
Cluster capacity now tracks real traffic, so the platform absorbs spikes without an engineer provisioning servers by hand, and quiet periods don't burn budget.
The Environment Is Reproducible
With every piece of infrastructure in version control, the team can review, audit, and rebuild the setup instead of relying on console clicks and tribal knowledge.
Automate Your Cloud Deployments and Scale With Confidence
DBB Software embeds DevOps engineers who codify your infrastructure and automate your path to production, so your team ships more often without giving up reliability.
Contact Us
"Most of our work starts with a 30-minute call where someone describes a product they're trying to ship and one part of the engineering picture they can't get around.
If that's where you are, let's set one up; I'll tell you straight whether we're the right fit.”
Mina Morkos
Business Development Manager
Want a similar outcome for your team?
Ask our AI assistant — it can pull related case studies, talk through the approach, and put you in touch with the team if you want a deeper conversation.
