What is the most difficult task you've ever worked on?

Situation: Migrating a multi-petabyte legacy data warehouse to a cloud-native lakehouse with zero downtime for 500+ daily users. Task: Achieve data consistency, performance parity, and seamless user transition without blocking business. Action: I designed a dual-write and dual-read strategy with automated reconciliation (row counts, checksums, sample validation). I built a query translation layer to map legacy SQL to Spark SQL, enabling gradual cutover. I migrated users in waves: low-impact reports first, then critical dashboards, then ad-hoc. Each wave had success criteria and rollback plans. I ran training, provided migration guides, and set up a support Slack channel. I used feature flags for instant rollback. Result: Migration completed in 6 months. 30% cost reduction, 50% faster P95 queries, zero major outages. The pattern was reused for two subsequent migrations.