Unum innovates across all layers of data-processing infrastructure:
- Storage: Persistent Data Structures, Compression and Zero-Copy Async IO,
- Compute: Matrix Multiplications, Graph Algorithms and Compiler Design,
- Modeling: Universal Encoder Architectures for Semantic Search.
For every layer we prepare a range of products, mostly written in C++ 20, CUDA 17 and Assembly for SIMD acceleration on x86 and Arm. Sounds interesting, check out what our storage engine is capable of.