A Sydney hedge fund wanted to explore satellite imagery and card-transaction datasets. We built an onboarding pipeline that kept the data auditable rather than a researcher's download folder.
Alternative data is valuable precisely because it is irregular. Satellite imagery arrives as tiles; card-transaction data arrives as weekly aggregates; neither fits neatly into a relational table. Without discipline, these datasets accumulate in individual researchers' download folders, and six months later no one can reproduce how a number was computed. The client wanted to explore alternative data without ending up in that state.
We built a landing zone that preserved the raw alternative data exactly as the vendor delivered it, plus a controlled metadata layer that recorded: vendor, licence terms, delivery date, coverage region, and the processing steps applied. We wrote a small set of reference processing jobs for the two datasets the client was actively exploring — aggregate weekly card indices and tile-level image summaries — but we deliberately did not over-build. The point was that anything a researcher does with the data leaves a trace in the metadata layer, so the next person can see how a number was derived.
The client's researchers could explore the new data without the data turning into an undocumented black box. When the research team later decided which alternative datasets were actually useful, the useful ones had clean metadata and the rest could be retired. We did not evaluate whether the alternative data produced alpha — that is the client's investment process, not ours. We made sure the data could be used reproducibly when they did.
Send your requirements. Reply includes a written scope and price.