atakan gül

ABOUT / 01

atakan gül

Cloud & Platform Engineer

Istanbul · open to roles

I build cloud infrastructure and the software around it. Running once is easy. I care what it costs, how it fails, and whether others can run it without summoning its author.

PRINCIPLES IN PRACTICE / 02

How I
work.

Six rules, in the order I apply them: measure, reduce, own, verify, ship, watch.

  1. 01

    Measure before changing.

    I profiled the Python jobs before changing anything, and the profile showed where the memory went. Memory dropped 40% and the OOM crashes stopped.

    Measured / 03
  2. 02

    Remove work before adding capacity.

    The Glue job reprocessed everything on every run. Watermarks took it from 3 hours to 7 minutes on the same resources.

    Measured / 03
  3. 03

    Give each responsibility one owner.

    Ansible builds the cluster and Helm ships the releases, and neither does the other one's job. CI can check releases against the cluster but can't deploy them.

    Design decisions §8 CI identity §4
  4. 04

    Verify the cause, not just the result.

    Every object in the 20 TB migration was ETag-checked. My CI probe also checks which policy denied its write, because a write can fail for other reasons.

    Measured / 03 The proof
  5. 05

    Ship exactly what you verified.

    One image gets scanned, pushed and signed, and it's only signed if its digest still matches the scanned one. Otherwise I'd be signing an image nobody scanned.

    Delivery §5
  6. 06

    Too little data is not a pass.

    When there isn't enough traffic to judge, my canary warns instead of passing. Production needs real or synthetic load first.

    Delivery §10

MEASURED WORK / 03

Measured
work.

The rules above, applied in production.

−40%

I profiled the Python analytics jobs that kept running out of memory. Memory use dropped 40% and the OOM crashes stopped.

Production · 2025 · Rule 01

3h → 7m

The Glue job reprocessed all its data on every run. Incremental watermarks fixed that, and it saves about $6K a year.

Production · 2025 · Rule 02

20 TB

100M+ objects moved with DataSync, with no downtime and every ETag checked.

Production · 2025 · Rule 04

TOOLS I BUILD / 04

Open source.

Agent tooling I build outside work.

axon

A Go runtime for LLM agents: agent loops, streaming, tool dispatch, sessions and context pruning.

GitHub

cortex

A terminal coding agent built on axon, with MCP support, persistent sessions, context checkpoints and multiple model providers.

GitHub

EXPERIENCE / 05

Where I’ve
worked.

Mar 2025 — Oct 2026

Siemens

Software Development Engineer · Cloud, Edge

  • Cloud infrastructure and data pipelines for industrial IoT predictive maintenance, in the IQ Fleet team (Digital Drive Train).
  • Moved 20 TB+ and 100M+ objects on AWS with no business downtime.
  • Redesigned an AWS Glue ETL job from 3 hours to 7 minutes, saving about $6K a year.
  • Profiled a pipeline that kept crashing the cluster with OOMs and cut its memory use by 40%.
  • Found a critical design bug that led to a system-wide data correction.
  • Added Wiz scanning of images and IaC before deploy, set up Loki for central logs, and built Athena-backed queries into Siemens Copilot.

Oct 2024 — Mar 2025

Orion Innovation

Software Engineer Intern

  • Network diagnostics platform for enterprise clients.
  • Fine-tuned an NER model on internal data (about 90% coverage), then built a PII pipeline around it with custom templates (98%).
  • Refactored legacy FastAPI services into controllers, services and repositories, and added JWT auth with role-based access.
  • Covered the new layers and auth flows with unit and integration tests (about 75% coverage).

Aug 2024 — Oct 2024

Nevotek

R&D DevOps Engineer Intern

  • Deployment infrastructure for a hotel tech platform. Learned that a good release is a boring one.
  • Set up blue-green deployments on Azure Web App slots, with zero downtime and instant rollback.
  • Moved infrastructure code from Terraform to OpenTofu modules for AWS and Azure.
  • Ran Jenkins fully declared with JCasC, and built Jenkins and Helm pipelines that keep deploy logic out of any one cloud.
  • Cut Docker images by 88% with two-stage builds, and made Trivy fail the build on critical vulnerabilities.

Nov 2023 — Mar 2024

Istanbul Bilgi University

Student Research Assistant

  • Cleaned research datasets and produced reproducible Python analyses and visual reports.

Dec 2022 — Jan 2024

Summarify

Software Engineer Intern

  • Data pipelines and backend for a social media analytics startup. My first DevOps role.
  • Migrated 100 GB from S3 to Azure Blob in stages, and benchmarked Elasticsearch across AWS compute and storage tiers, including local NVMe.
  • Built RAG document Q&A with LangChain and FAISS, and cut report generation from 60 to 5 minutes with Flask and LLMs.
  • Built REST APIs over 50 GB+ of social media data and integrated iyzico subscription payments.

EDUCATION & CERTIFICATION / 06

Education &
certification.

Istanbul Bilgi University

B.Sc. Computer Engineering

2022 – 2026

Full scholarship (top 0.3% nationally)

AWS Certified Solutions Architect — Associate

Amazon Web Services

Issued July 2026

View credential