Alperen
Adalar

At Ecolarium, infrastructure and monitoring are mine. Every production service stayed above 99.9% uptime over the last 6 months; this is what I run today.

Kubernetes

3+ kubeadm clusters in production, under 10 nodes in total. Releases and rollbacks go through Helm.

3+ clusters

PostgreSQL

3-node HA with streaming replication. I upgraded a public institution's PostgreSQL from 14 to 18. Backups run with pg_dump and we rehearse restores regularly.

1 TB+

Oracle → PG

Moved the same institution's Oracle database to PostgreSQL with ora2pg: 606 tables, about 3 billion rows, 142 sequences and part of the views and functions.

~650 GB

Observability

OpenTelemetry into self-hosted SigNoz, with ClickHouse on its own disk. Retention is set separately per data type and log category, sized to the disk budget.

254 spans

Alerting

5 alert rules in SigNoz. Prometheus, Alertmanager and blackbox exporter on a second product, with 4 incident playbooks.

5 rules · 4 playbooks

Servers

10–30 servers and VMs, plus 1–3 GPU servers I run myself. FortiGate / pfSense, Keycloak SSO and Nginx with TLS at the edge.

10–30

CI/CD

Self-hosted GitLab CE with 4 runners, one on Windows. 459 pipelines ran in the last 90 days.

168 repos

Kafka

Live device and dashboard updates flow through Kafka and Django Channels.

1,000+ events/s

1. Jul 2025 – present

   Türkiye

   DevOps & Platform Engineer

   Ecolarium Dijital Teknolojiler

   • I run the infrastructure for Vitalert Vision and our other products: 3+ production Kubernetes clusters (kubeadm), 10+ Helm charts for releases and rollbacks, 3-node PostgreSQL HA with streaming replication, 10–30 servers and 1–3 GPU servers. Every production service stayed above 99.9% uptime over the last 6 months.
   • Instrumented Django, Celery, PostgreSQL, Redis and LLM calls with OpenTelemetry (254 spans). Set up SigNoz on ClickHouse on our own server, linked logs to traces, and wrote 3 dashboards and 5 alert rules.
   • Set up Prometheus, Alertmanager and blackbox exporter for a second product, tested the alert rules with promtool, and wrote 4 incident playbooks.
   • Upgraded a public institution's PostgreSQL from 14 to 18 (1 TB+ of data) and brought back a replica that had been stuck since May. Traced 152 slow commits to a nightly VACUUM FULL.
   • Completed the same institution's Oracle-to-PostgreSQL migration with ora2pg: 606 tables, about 3 billion rows (~650 GB), 142 sequences and part of the views and functions.
   • Upgraded GitLab CE (168 repos, 4 runners) and moved its host from CentOS 7 to Rocky 9, then wrote a usage-based comparison with Gitea and Forgejo.
   • Built Kafka and Django Channels pipelines that handle 1,000+ telemetry events per second.
   • On the backend side: a multi-tenant ESG-reporting SaaS on Django and django-tenants.

   • Kubernetes
   • Helm
   • PostgreSQL
   • OpenTelemetry
   • SigNoz
   • Prometheus
   • GitLab CI
   • Kafka
   • Django

2. Jan 2024 – Dec 2025

   Konya

   Student Researcher

   Konya Technical University · TÜBİTAK 1711 R&D Project

   • Band-saw predictive maintenance project with İMAŞ (TÜBİTAK TEYDEB 3237008), also my graduation project. I owned the camera-based saw analysis and the machine kiosk.
   • Trained YOLO and RT-DETR models for broken-tooth and crack detection (tooth model: mAP50 0.995 on validation) and wrote real-time anomaly detection for the sensor data (z-score / IQR).

   • Python
   • PyTorch
   • YOLO
   • RT-DETR
   • PySide2
   • Modbus
   • ThingsBoard

3. Jul 2023 – Dec 2024

   İstanbul

   Data Engineer / DevOps Engineer

   EVAM

   • Set up Zabbix for 50+ network devices, servers and applications and wrote the alert rules myself.
   • Built and ran AWS infrastructure (EC2, S3, RDS) for production data and application workloads.
   • Built Jenkins CI/CD pipelines for 10+ containerized services on Kubernetes and Rancher.
   • Wrote 30+ Airflow DAGs for ETL and moved most of the manual jobs onto them.

   • Zabbix
   • AWS
   • Kubernetes
   • Rancher
   • Jenkins
   • Airflow
   • Keycloak
   • Superset

4. Oct 2021 – Jul 2023

   Konya

   Software Developer

   IMAS Machinery Industry

   • Ran ThingsBoard server and edge clusters for 20+ industrial devices, and Kafka for machine-to-analytics streaming.
   • Connected PLCs and single-board computers over RPC, TCP/IP and Modbus (RTU/TCP).
   • Wrote Python pipelines for kHz-rate sensor data from 30+ channels per machine.
   • Built scikit-learn anomaly-detection models for predictive maintenance.

   • ThingsBoard
   • Kafka
   • Python
   • Modbus
   • scikit-learn
   • Siemens S7

5. Dec 2021 – Feb 2023

   Konya

   Software Developer · Part-time

   Konya Science Center

   • Looked after the Linux servers and Docker services that ran the exhibits 24/7.
   • Shipped augmented-reality applications that visitors used every day.
   • Built native Android applications for the exhibits.

   • Linux
   • Docker
   • Android
   • AR

Open-Source MCP Servers for Infrastructure

Model Context Protocol servers I wrote so an LLM can operate Active Directory, the network, Proxmox and FortiGate in plain language. Built with Python, FastMCP, Pydantic and Docker; they work with Claude, Cursor and other MCP clients.

Open to DevOps and platform positions, and to infrastructure problems worth thinking about. Email is fastest: