amawta
Portfolio

Eigen Suite

Our work on embeddings, KV-cache, and model-weight structure led to compression and optimization tools. It is not the main offer; it is evidence of technical capability.

Data compression

EigenDB

Vector embedding compression up to 40× in internal benchmarks.

40×compression
40× Compression
>95% Precision
Zero Training
Learn more
Memory optimization

EigenKV

Reduces LLM KV-cache by 1.7× with <1% quality loss.

1.7×reduction
1.7× Reduction
<1% Loss
Drop-in
Learn more
Inference optimization

EigenWeights

Accelerates inference up to 30% by optimizing MLP layers.

30%faster
30% Faster
Plug & Play
No Retraining
Learn more

See how research results become practical tools.