NYC Taxi Data Warehouse (Snowflake + dbt)

Production-style mini-warehouse on the NYC Yellow Taxi dataset.
Raw Parquet ➜ Snowflake RAW ➜ ANALYTICS views ➜ dbt models, tests & lineage.

• Warehouse: Snowflake (X-Small warehouse)
• Modeling: dbt (Snowflake adapter)
• Orchestration/Load: simple stage → SQL inserts (safe timestamp parsing) + optional proc/task
• Docs: dbt docs (catalog + lineage)

---

✨ Highlights

• Clean ingestion: Parquet files uploaded to an internal stage, loaded with idempotent SQL.
• Safe parsing: TRY_TO_TIMESTAMP_NTZ() prevents bad rows from breaking loads.
• Curated layer: ANALYTICS views for BI (daily KPIs, by payment type, top pickup zones).
• Quality: dbt tests (not_null, accepted_values) + searchable docs & lineage.
• Cost control: X-Small warehouse, AUTO_SUSPEND=60, AUTO_RESUME=TRUE.

---

🗺️ Architecture

flowchart TD
  A["NYC TLC Parquet (Yellow trips)"] --> B["@YELLOW_STAGE (internal stage)"]
  B --> C["RAW.YELLOW_TRIPS (table)"]

  C --> D["ANALYTICS.V_DAILY_METRICS (view)"]
  C --> E["ANALYTICS.V_DAILY_BY_PAYMENT (view)"]
  C --> F["ANALYTICS.V_TOP_PU_ZONES (view)"]

  %% dbt lineage (optional, for docs)
  C -. "dbt source" .-> S["nyctaxi.YELLOW_TRIPS"]
  S --> T["stg_yellow_trips (dbt view)"]
  T --> M["fact_daily_metrics (dbt table)"]
  M --> X["Exposure: nyc_taxi_daily_metrics"]

📸 Screenshots

Stage files

Monthly counts

dbt lineage — staging

dbt lineage — fact