Andrés Estupiñán

Results & Impact

• Run platform engineering across six AWS projects for an automotive group, entirely in Terraform pipelines running on GitLab/GitHub. Supported the migration of the workloads from GitLab to GitHub and built a unified Grafana dashboard aggregating metrics from all six, including a Bedrock AI RAG system. Also delivered security-posture work with the client security team to bring two projects from DEV to PROD.
• Built an end-to-end GenAI data platform for a premium automotive manufacturer: a Glue and PySpark ETL pipeline consolidating production-line data into an Apache Iceberg lakehouse, fronted by an agentic assistant on Amazon Bedrock querying Athena through a Converse API. Production-line staff who write no code now ask in natural language instead of queueing behind the one engineer who could query the underlying Oracle database. The assistant returns the SQL it generated, so a harder question can be passed straight to an engineer to extend. Engagement delivered entirely in German.
• Migrated 42 of a logistics and shipping enterprise's most critical AWS accounts to a standardized Terragrunt and GitLab CI/CD model in a KRITIS-regulated environment, including the Shared Transit Gateway that all ~350 accounts in the estate route through.
• Ran DevOps for a commercial-vehicle manufacturer's hub-and-spoke data lake and RAG platform: AWS CDK and GitLab CI/CD, Lake Formation cross-account data sharing, and clearing every AWS Security Hub finding on the hosting account of a vLLM-based Retrieval-Augmented Generation (RAG) workload.

Results & Impact

• Designed and deployed secure AWS middleware for querying Starburst data: a Flask API handling OAuth2 authentication over Trino with mTLS and IAM, running on ECS Fargate behind an Application Load Balancer, with ECR, VPC endpoints and KMS-encrypted Secrets Manager, all provisioned in Terraform. Delivered query-indexing recommendations and load testing for large datasets.
• Migrated a manually managed Kubernetes cluster to Terraform and implemented GitOps CI/CD with GitHub Actions and ArgoCD.

Results & Impact

• Boehringer Ingelheim, Dataland (approx. one year). GxP-regulated enterprise data platform on AWS, later published by AWS as a customer case study. Deployed AWS Professional Services' data.all data marketplace in AWS CDK and Python: inter-region VPC peering, Amazon SageMaker, Redshift Serverless and MLflow.
• In this project also led the AWS Professional Services delivery team for the final three months of the engagement: client sprint reviews, task prioritisation across the AWS team, and the technical bridge between AWS and the client on the project's progress.
• Covestro Deutschland AG. Data Architect on the Serverless Data Lake Framework (SDLF), migrating Tableau prep-flows to AWS Glue (PySpark). AWS later published the DataZone data mesh built on that foundation as a customer case study.
• Built a GitLab pipeline for an ML-powered document and order-processing workload for a company in the Chemicals Manufacturing sector.
• acierto.com (Spain). Built a Terraform observability stack for EKS on Fargate: ADOT, kube metrics and events, X-Ray, and CloudWatch dashboards generated in Python. Presented the project outcomes on site in Barcelona at the clients' biannual on-site meeting.
• Built Terraform pipelines for EKS with security scanning (TFSec) in CodeBuild for a mobile bank in Abu Dhabi, and automated cluster tooling installation with Helm charts.
• Built multi-account VPC and Transit Gateway automation with advanced AWS CDK configurations for a telecommunications client in Israel.

Results & Impact

• Implemented an automated AMI customization pipeline for Windows Server with Packer and GitLab, after solving the Packer-to-Windows-EC2 connection problem that had blocked this automation for the team. This served as a foundation for an article written later for the AWS Blog Post.
• Responsible for the production and new releases via GitLab of the AWS backend (EC2, Auto Scaling, ELB, RDS, CloudWatch) of the point-of-sale system running on mobile devices across the entire ICE high-speed train fleet.
• Supported release roll-outs across test and production environments via GitLab for different projects. Engagement delivered in German and participated in on-call rotation.