DriftFixer

DriftFixer monitors production data pipelines and warehouse tables for “silent breakage”: gradual shifts in distributions, null-rate creep, cardinality explosions, and join-key decay that don’t trigger traditional job-failure alerts. Connect Snowflake/BigQuery/Redshift and a scheduler (Airflow/dbt) to auto-profile selected tables, set lightweight baselines, and generate drift scores per column and per dataset. Alerts include plain-English impact summaries (e.g., which metrics will likely move) and a short list of probable upstream causes (new source version, late-arriving data, schema change, filter regression). It’s a web app with an AI-assisted explanation layer, but the core detection is deterministic and auditable. The product is intentionally narrow: drift detection and triage, not a full observability suite. Expect some false positives early; the value is reducing time-to-diagnosis when data “looks fine” but business numbers are wrong.