Composable foundations for genomic data.
Building a shared, Arrow-native substrate that lets genomic data flow natively through the modern data stack. Declarative, streaming, and portable, from cloud storage to GPU.
the idea
Declare. Stop parsing.
Genomic data is, at heart, data on a coordinate system — interval-based features located along a reference, with a few domain quirks like strandedness. Yet the tooling around it is fragmented and welded to file formats, so most effort goes into parsing, copying, and shuttling data between programs.
CompoSeq flips the model: bring genomic data into general-purpose engines that already solve joins, streaming, and cloud storage — then layer on the few genuinely genomic primitives: interval operators, coverage and pileup rasterization, and coordinate-aware tensor loading. Generic work is optimized once, in shared infrastructure; specialized kernels stay modular and reusable over Apache Arrow.
projects
Maintained libraries working together.
team