2 min lesson
Separate environments to bound blast radius
Think through this situation: "A product team wants to start using developer code and usage data to improve an agent feature. How do you say yes responsibly rather than blocking or rubber-stamping it?" Give the practical answer in plain words.
Step 1 of 3
Separate environments to bound blast radiustraining vs. analytics vs. raw
Interactive diagram. Tab through its regions; each focused region shows its detail in the panel below.
Separation by design: a leak in analytics cannot reach raw unmasked PII and the agent feed lives apart from both.
Learn more
Full explanation
Retention, deletion and the right to be forgotten
Retention, deletion and the right to be forgottendeletion across a lakehouse is genuinely hard
A deletion request is easy to honor in a single transactional database and hard across a lakehouse where one user's events are scattered through partitioned Parquet, copied into silver, aggregated into gold and maybe cached. You need a plan: a way to find every copy via lineage, Delta's delete/vacuum to physically remove rows and retention policies that age data out so you're not hoarding what you don't need.
"We deleted the row" is incomplete if the data was also pre-aggregated into gold, copied to a training set or sitting in a cached extract. Use lineage to enumerate every copy, delete and vacuum the underlying files (a logical delete that leaves old Delta versions isn't gone) and confirm the replicas. Deletion that only hits the obvious table is the kind of gap an audit finds for you.