David Moscatelli, founder and CEO of Go Abacus, says organizations in regulated industries are increasingly wary of using public AI providers due to data privacy concerns. Banks, credit unions, and hospitals are seeking ways to leverage AI locally, without sending sensitive data to the cloud.
Go Abacus aims to provide AI infrastructure specifically designed for these regulated sectors. Moscatelli notes that the preference for local AI solutions is also driven by cost control, as institutions want a fixed price rather than variable cloud costs.
The company's flagship product, the Go One device, enables on-premises AI by connecting to employee PCs and preconfiguring software. According to Moscatelli, a single device can support up to 2,000 concurrent users, and scalability is achieved by daisy-chaining multiple units.
Moscatelli demystifies large language models, describing them as essentially CSV files with weights, combined with software to run those weights. The Go One OS uses a collection of specialized models trained on client-specific data for deterministic tasks. Models are trained during off-hours, with updated weights sent back to clients nightly.