Leave request
Presentation
Burger logoBurger logo
Case study · Platform modernization + AI operations

From manual deploys to AI-supervised operations

How a high-traffic reviews aggregator moved from docker-compose to Kubernetes with unified CI/CD, then handed daily operations to Metatron, Digital Care’s AI control plane.

Industry Consumer webWorkload Reviews aggregatorStack Kubernetes + Metatron
An AI node supervising an orderly field of glowing microservices.
The situation

The client runs a high-traffic reviews aggregator built as microservices. All containers ran through docker-compose and every deployment was manual: no isolated environments, no repeatable release process, and operations that depended entirely on individual engineers.

The goal

Move the platform to Kubernetes as the single orchestrator, unify CI/CD for all applications, and then take the operational load off the team for good.

Before
Manual
every deploy, by hand

A microservice architecture running through docker-compose. Deployments were manual and error-prone, there were no isolated environments for testing, and every release was a small project of its own.

After
AI on call
Metatron watches the platform

Kubernetes with three isolated environments and automated GitLab pipelines: deploys ship without downtime. And since Metatron joined as the control plane, every alert is investigated automatically, with root cause posted to Slack in minutes.

What we built

Modernized the platform. Then put an AI on call.

01

Kubernetes migration

Deployed a new Kubernetes cluster and moved every microservice off docker-compose onto declaratively described workloads.

02

Three environments

Dev, stage and prod: changes are tested and rehearsed before they ever reach users.

03

CI/CD on GitLab

Build and deploy pipelines in DinD mode: extra isolation, convenient scaling, automatic releases without downtime.

04

Monitoring

Prometheus + Grafana across all services: resource consumption and availability visible per service.

05

Metatron as the control plane

Our AI agent watches the whole platform: every Alertmanager alert picked up and investigated automatically, evidence-backed root cause in Slack in minutes. On-call engineers freed from routine investigation.

#reviews-oncallSlack
MetatronAPP07:13
Incident resolved: reviews-api
ROOT CAUSEOOMKilled · memory limit 512Mi exceeded
EVIDENCE4 signals correlated · logs, metrics, events
RECOMMENDRaise limit to 1Gi · GitOps PR #217
TIME TO RCA3m 41s
Read-only investigation, awaiting human approval
Technologies
GitLabKubernetesHelmPrometheusGrafanaTerraformAnsibleMetatron
Results
0 downtime
automatic deploys via GitLab CI/CD
3 envs
dev, stage and prod, isolated and reproducible
100%
of alerts investigated by Metatron
Minutes
from alert to root cause in Slack
The product behind it
Metatron: the AI infrastructure control plane
Next case
Finland