Abstract: A system and method for distributed exchange and transformation of structured data with reproducibility and complete provenance are described. Data enters the system through the plurality of data publishers that share data in the form of append-only historical logs of events. Plurality of parties can then collaborate on improving and enriching data by defining the transformation rules, forming a potentially distributed computational graph. Data from any stage of such graph can be accessed by plurality of consumers. Stream processing techniques are used to minimize the propagation latency of data through the graph. Bitemporal data modelling and determinism of transformations enable reproducibility and verifiability of the results. Every event that influences how data looks like is tracked in the temporal metadata which enables provenance and dataset evolution.