Build the pipelines behind datastore: decoding years of on-chain history into clean, versioned Parquet that researchers can trust.
datastore sells something unusual: files, not API access. Customers download decoded on-chain history as Parquet and run their own queries. That only works if the data is actually right, which makes correctness, lineage and reproducibility the product.
You will build and run the pipelines that decode Solana and Hyperliquid history at scale: backfills over billions of rows, schema design, checksums, manifests and the quality checks that let a quant trust a file they did not produce themselves.
What you will do Design and run large-scale decoding and backfill pipelines
Model typed schemas for instructions, events and state across dozens of protocols
Build validation that catches bad data before customers do
Keep versioning, checksums, manifests and lineage docs accurate on every delivery
Tune storage layout and partitioning so files query fast in DuckDB and Polars
Add new protocols and chains to the catalogue
Support customer questions about schemas and coverage
What we are looking for Experience building production data pipelines at meaningful scale
Robust SQL plus one of Python, Rust or Go
Real familiarity with columnar formats,



ideally Parquet, and query engines like DuckDB, Polars or Spark
Care for data correctness: testing, validation and reconciliation
Comfort owning pipelines in production, including when they break at night
Able to work from our Toronto office part of the week
Nice to have Experience with blockchain data or other messy, high-volume event streams
Familiarity with warehouse ecosystems your customers use, like Snowflake, BigQuery or ClickHouse
Background in quantitative research support or backtesting infrastructure
Experience with orchestration tools and with knowing when a cron job is enough
Who you are You think an unverified number is worse than no number
You write pipelines you would be happy to debug at 2am, so they rarely need it
You like schemas that make the next person's query obvious
You get satisfaction from a backfill that reconciles to the last row
How we hire 1 Intro call with an engineer, about 30 minutes
2 Short take-home assignment working with a real decoded dataset
3 Technical conversation about your assignment and pipelines you have run
4 Conversation with the founders
5 Offer

#J-18808-Ljbffr

📌 Data Engineer (Ontario)
🏢 CLR3
📍 Ontario