Required skills
- Amazon SageMaker
- MLOps
- Amazon Web Services
- Infrastructure as Code
- AWS CloudFormation
- Pulumi
- Amazon Neptune
- Gremlin
- GraphStorm
- Deep Graph Library (DGL)
- CI/CD
- GitLab CI
Languages: English
Job description
Project overview
DevOps / MLOps Engineer (AWS, GitLab CI, Pulumi) – 90% remote, 3-year engagement (01.01.2027 – 31.12.2029, approx. 220 days/year).
Project goal:
For a large German mobility corporation (IT subsidiary), we're looking for up to 3 freelance DevOps / MLOps Engineers to run, automate and keep improving the cloud, CI/CD and MLOps platform behind its production ML solutions. These include fraud detection in online ticket sales and the optimisation of operational review processes. You'll work in an agile team on AWS.
Responsibilities
- Run and develop GitLab CI pipelines across a multi-repository landscape: central CI templates, reusable pipeline components, cross-project integration, and deployments per environment
- Automate container builds with Artifactory, including vulnerability scanning, image versioning and delivery to AWS ECR for separate test and production environments
- Design and build AWS infrastructure as code (CloudFormation, Pulumi): reusable components, multi-account architectures, configuration and change management
- Automate MLOps deployments on AWS SageMaker: Model Registry, approval workflows, EventBridge, Lambda, and automated rollout of production inference endpoints via GitLab CI
- Standardise release processes across AWS accounts and environments, including approval gates, acceptance and shadow deployments, and controlled rollbacks
- Build and run internal web applications and backend services (REST APIs, admin and monitoring UIs, authentication and authorisation)
- Manage dependencies and releases with uv, private Python registries on Artifactory, and Renovate
- Operate and optimise RDS/Aurora, PostgreSQL, Amazon Neptune, Kafka/MSK, ECS Fargate, Lambda, Step Functions and private REST endpoints
- Provide infrastructure for graph-based data and ML workloads (Neptune/Gremlin, GraphStorm/DGL)
- Set up monitoring and observability with CloudWatch, X-Ray, SNS and Grafana, including alerting and incident analysis
- Automate security and compliance: least-privilege IAM, Secrets Manager, KMS encryption and key rotation, automated security checks
- Automate data retention and deletion (S3 tagging, lifecycle policies, Step Functions) and hybrid data transfers between Azure Blob Storage and AWS S3
- Build infrastructure for AI agent systems (MCP servers, CLI tool integration, observability with Langfuse)
- Incident handling, performance tuning and technical documentation (service manifests, deployment configurations)
Requirements
- 3+ years with AWS SageMaker and production MLOps (Model Registry, pipelines, endpoint deployments, automated approval workflows)
- Solid hands-on AWS architecture experience: IAM roles and policies, VPC network design, KMS, multi-account environments
- Production infrastructure as code with AWS CloudFormation or Pulumi
- Hands-on experience with graph technologies (Amazon Neptune/Gremlin or GraphStorm/DGL), including running graph-based data or ML workloads
- CI/CD and deployment processes in regulated environments (approval procedures, traceability, environment separation, security and compliance)
- English C1 (no German required)
Nice to have
Timeframe: 01.01.2027 – 31.12.2029, approx. 220 days per year.
Working model: 90% remote from the EU or Germany; approx. 20 onsite workshop days per year in Frankfurt am Main.
Expected outcome: A stable, secure and fully automated cloud and MLOps platform on AWS that reliably delivers and runs the client's production ML and web applications.
- Pulumi with Python, including reusable IaC components and libraries
- Full-stack web development (REST APIs, authentication, automated cloud deployments), 3+ years
- Event streaming with Apache Kafka / AWS MSK
- Hybrid data transfers between Azure Blob Storage and AWS S3
- AI agent systems, MCP servers, Langfuse observability
Project details
- Start: 2027-01
- Remote: hybrid
- Location: Frankfurt am Main, Deutschland, DE
- Languages: English
Indexed by 9am's meta-search. You are a job provider? Claim or manage this listing
Found on FreelancerMap DE