Murat Can Üste
Sr. Software Engineer - DevOps/MLOps
Munich, Germany
About
Senior Software Engineer building the internal platforms and developer tools that let engineering teams ship faster and break less. I write Kubernetes operators, CLI tools, and CI/CD systems in Rust and Python — and I genuinely enjoy making infrastructure disappear so developers can focus on their actual work. Before DevOps, I spent years in autonomous driving research at TU Munich, publishing on motion planning, imitation learning, and graph neural networks for traffic scenario analysis. That research background gave me a deep appreciation for rigorous systems thinking, and a slight obsession with making complex things feel simple.
Experience
Tensordyne
Sr. Software Engineer - DevOps/MLOps
Mar 2024 - Present | Munich, Germany
- Developed a Kubernetes Operator in Rust for deploying LLM inference servers on specialized hardware, with pod lifecycle reconciliation, admission webhooks, and Python bindings for ML engineers.
- Designed and deployed a multi-cluster, multi-region Cloud Development Environment on GKE using Coder, supporting GPU workloads with reusable Terraform templates and IAP-based authentication.
- Deployed Argo Workflows as a scalable CI backend to replace Jenkins, integrating FluentBit for log collection and Grafana/Loki for monitoring.
- Co-developed an in-house CI engine on Argo Workflows enabling developers to define dynamic CI/CD pipelines via HCL configurations.
- Built reusable Terraform modules for GKE clusters, node pools, service accounts, DNS, and local NVMe SSD RAID provisioning, standardizing infrastructure across environments.
- Unified Python linting and type-checking across a 7+ project monorepo by migrating to Ruff, Pyright, and uv, enabling security, bug detection, and strict type-checking rules.
- Built a cross-platform Rust CLI for authenticating with internal developer platform tools such as Coder, Jenkins, Argo Workflows, and Flyte with identity-based access controls.
- Built a Rust replacement for gsutil that cut CI artifact handling time by 4x and reduced Docker image sizes by 20x.
- Supported the monorepo consolidation, writing migration documentation and providing hands-on developer support for new tooling and quality checks.
- Maintained a GPU cluster usage dashboard for Flyte-based ML workloads, enabling resource optimization across teams.
PreciBake
DevOps Engineer
Sep 2022 - Feb 2024 | Munich, Germany
- Established policies for SDLC, artifacts, and container management based on Trunk-Based Development, increasing release and deployment frequency.
- Developed GitLab CI/CD templates for organization-wide automation of builds, tests, and containerization processes.
- Developed scripts, CLI tools, and containers for CI pipelines and multi-arch builds.
- Developed Python project templates with best practices, improving code quality and developer workflow.
- Developed Ansible Playbooks for deploying to 100+ edge devices and servers via Ansible AWX.
- Managed on-prem Kubernetes clusters (10+ nodes, multi-zone) using Rancher Stack, Traefik, Kyverno, and others.
- Automated and defined policies for deployments to Kubernetes clusters using Helm and Gitlab CI/CD.
PreciBake
Software Engineer
May 2022 - Aug 2022 | Munich, Germany
- Optimized build and containerization for video streaming and object detection software, reducing container size from 6GB to 700MB and expanding deployments to AMD64 and ARM64 architectures.
- Automated the packaging and containerization using GitLab CI/CD, leading to increased developer productivity and reliable builds and deployments.
Technical University of Munich - Chair of Robotics
Software Engineer
Feb 2019 - Apr 2022 | Munich, Germany
- Led a 4-member team in developing the CommonRoad Platform for benchmarking autonomous vehicle motion planning algorithms.
- Developed CommonRoad Platform with Python, Django, React; used PostgreSQL, MinIO storage, and integrated RabbitMQ and Celery for distributed task processing.
- Implemented CommonRoad evaluation metrics to measure the performance of autonomous vehicle trajectories based on safety, efficiency, and other parameters.
- Developed the feasibility checker module of the CommonRoad Drivability Checker Toolbox to simulate and assess the feasibility of automated vehicle trajectories.
- Designed CLI tool for platform interaction and streamlined motion planning algorithm development and evaluation in sandboxed environments with Docker.
- Automated containerization, testing, and deployments to LRZ Cloud and self-hosted servers via GitLab CI/CD and Ansible, increasing deployment frequency.
- Hosted institute-wide and global competitions on the platform, evaluating and visualizing 6,400 scenarios and 320,000 trajectories.
Linovi
Software Engineer
Nov 2017 - Oct 2018 | Izmir, Turkey
- Collaborated in a 4-member team to design and integrate platforms like Jira, GitHub, and Zendesk with Google Chat as chatbots, serving over 3,000 unique daily users at launch for G Suite customers.
- Utilized JavaScript and Scala to ensure backend responsiveness, achieving under 200ms end-to-end response times.
- Containerized chatbots for deployment on GCP, utilizing services like Functions, Spanner, and Memorystore (Redis).
Technische Hochschule Mittelhessen - WebMedia & iCampus
Software Engineer (Internship)
Jul 2016 - Sep 2016 | Gießen, Germany
- Developed a Jenkins job status monitor using JavaScript and Node.js for real-time Jenkins API data retrieval.
- Partnered with the iCampus group to craft Joomla CMS extensions.
Ege University - Institute on Drug Abuse, Toxicology and Pharmaceutical Science
Software Engineer (Student Assistant)
Oct 2014 - Jun 2016 | Izmir, Turkey
- Collaborated with a 3-member team to create an Inventory Management System using PHP and MySQL, optimizing monthly reporting.
Education
Technical University of Munich
M.Sc. Informatics
Sep 2018 - Apr 2022 | Munich, Germany
Ege University
B.Sc. Computer Engineering
Sep 2013 - Jun 2017 | Izmir, Turkey
Publications
CommonRoad Drivability Checker: Simplifying the Development and Validation of Motion Planning Algorithms
C. Pek, V. Rusinov, S. Manzinger, M.C. Üste, M. Althoff — IEEE Intelligent Vehicles Symposium (IV), 2020
2020
Projects
Generation of Naturalistic Traffic Rule Violations Using Imitation Learning
Master's Thesis at TU Munich — Using GAIL to generate naturalistic traffic rule violations for autonomous driving testing.
Assessing the Similarity of Traffic Scenarios Using Graph Neural Networks
Research project at TU Munich — Using DDGK to learn graph-based similarity embeddings for traffic scenarios.
Motion Planning with Temporal Logic Specifications
Seminar work at TU Munich — Reviewing integrated task and motion planning with temporal logics for robotics.
Motion Planning Feasibility Checker
CommonRoad Drivability Checker module for simulation and feasibility evaluation of automated vehicle trajectories.
Clustering Similar Traffic Scenarios
Research project at TU Munich — Using unsupervised Random Forest to automatically categorize traffic scenarios.
Emotion Analysis of Meetings Using Facial Expression Recognition
Bachelor's Thesis at Ege University — Real-time facial expression recognition for speaker feedback.
Skills
Programming
Databases
Frameworks
DevOps
ML
Tools
Languages
Turkish — Native
English — C2
German — B1