Problem
The machine holding sequencing data often can’t host a large reference database, and medical data can’t always leave the site. SKiMet splits the work: a server holds the index, clients hold the reads.
What I built
- The gRPC protocol, connection lifecycle, and streaming interface. Reads stream up, classifications stream back.
- A server pipeline on Intel TBB that overlaps I/O, decompression, and lookup, so steady traffic doesn’t stall on any one stage.
The privacy-preserving encryption layer is planned future work. I haven’t worked on it.
Stack
C++23 · gRPC · Intel TBB · CMake