Data & Automation Engineer · Working Student
Data Engineering at Otto Group one.O
When two reports disagree, which one is wrong, and where did it go wrong?
Schematic illustration · not project data
- Role
- Data & Automation Engineer, Working Student
- Where
- Otto Group one.O (Otto Group)
- Period
- Feb 2023 – Dec 2025
- Stack
- Oracle, SQL, Python, Power BI, Power Automate, Elastic Stack, REST APIs, GCP, Linux
Overview
An analysis is only as reliable as the data definitions under it.
For nearly three years, alongside my studies, I worked on the data that a team's recurring reporting was built on. Most of that work happens before anyone opens a dashboard: in views, scheduled jobs, definitions, and checks.
It is also where I formed the habit that now drives my research. When a figure looks wrong, or looks too good, the useful question is which step produced it.
Work
- 01
Reporting pipelines
Built and maintained the SQL views and scheduled data jobs on Oracle behind the team's recurring reporting, and checked results against business expectations before they shipped.
- 02
Tracing errors to their source
Traced wrong figures in downstream dashboards back to the view or job that produced them, and fixed the cause rather than the number.
- 03
Consistent definitions
Rebuilt the team's data definitions and technical documentation from scratch, which resolved inconsistencies between reports that had shown conflicting figures.
- 04
Automated archiving
Replaced a manual filing process with a REST API-based archiving service that ingests documents and extracts their metadata automatically.
- 05
Dashboards
Built Power BI dashboards for business stakeholders and walked them through what the data could and could not tell them.
- 06
Automation
Automated recurring data processing and monitoring tasks in Python and Power Automate.
- 07
Monitoring
Helped design the company-wide Elastic Stack monitoring setup and rolled it out, covering log processing and alerting.
- 08
Operations
Administered Linux and Windows servers, provisioned cloud databases on GCP, and ran server-side performance tests.
Carried forward
My B.Sc. thesis, a framework for matching heterogeneous product datasets, was written with one.O during this time. It is described under Applied AI & LLMs.