On-chain data engineering. I build the whole path rather than querying someone else's index — raw logs off RPC nodes, decoded into BigQuery, priced hourly, published as static dashboards. Currently Uniswap v3 and v4 on Robinhood Chain.
live · static JSON
things the data said
python · bigquery · sql
Ingestion in Python across several RPC providers at once, block-range sharded, with per-day correctness checks against a second provider before anything is loaded.
BigQuery, partitioned by day and clustered on the columns queries actually filter, with a partition filter required on every raw table so a careless query is rejected at $0 rather than scanning a terabyte.
Transforms are incremental and keyed, so a re-run replaces rather than duplicates, and every window is fenced so it cannot touch history it wasn't asked to look at. Every query is dry-run and priced before it spends.
Aggregates export to static JSON on object storage, so the warehouse is never in the request path.