Research at Nova

We present a novel approach to training large language models that reduces computational requirements by 60% while maintaining model performance. Our method is particularly suited for enterprise applications where data efficiency is crucial.

A comprehensive framework for implementing secure multi-party computation in distributed AI systems, enabling multiple parties to jointly compute AI models while keeping their input data private.

An innovative approach to neural architecture search that specifically targets resource-constrained enterprise environments, achieving state-of-the-art performance with minimal computational overhead.