Ad technology integrations between enterprise customers and Google Cloud, plus analytical ETL pipelines that aggregate operational and advertising data into dashboards the ad-tech teams actually read. Tracing data-flow issues through product and integration layers until the cause stops being a mystery.
GraphQL client libraries in Go and Python, used across internal teams to pull from shared APIs. Polyglot microservices in Java, Python, and Go behind them. Profiling and load testing on the hot paths; Terraform and Docker on the deployment side. A crash-course in code review as craft.
Big-data pipelines inside BMW's data estate. Hadoop MapReduce and PySpark on the batch side, Airflow DAGs holding the orchestration together, and Go worker services on a pub/sub event bus for the streaming half. Learned that most "data problems" are really ownership problems with a schema on top.
Payment microservices in Python and Go on RabbitMQ and Redis, plus an SDLC analytics pipeline that fed Jira and Git metrics through Elasticsearch into Kibana dashboards. Stripe integrations on the money side. The year that convinced me architecture is mostly the art of deciding what not to build.
The working kit accumulated across the four roles above. The emphasised items are the ones I reach for by default.