Index / 03
Technology
The modular platform behind every VIVERRA product — models, knowledge, tools and compute.
Platform model
A modular field of models, knowledge, tools and compute.
The platform model separates reusable intelligence services from product-specific experiences. This keeps interfaces focused while allowing infrastructure and capabilities to evolve underneath them.
Product surfaces
Purpose-built web, desktop and private experiences with distinct navigation and product language. Each surface is generated from one product manifest instead of being rebuilt from scratch.
Agent systems
Specialized assistants, workflows, tools and structured actions composed for each use case. Outputs are governed by strict Zod schemas, so agents complete precise JSON workflows deterministically instead of returning free-form guesses.
Knowledge layer
Document libraries, retrieval-augmented answers and project context organized around explicit access boundaries, with provenance and a consent ledger for auditable data use.
Execution plane
Responsive product control runs on a serverless control plane; longer processes, terminals, sandboxes and GPU workloads run on dedicated or external compute with provider abstraction and hard spending caps.
Local-first inference
Models are compiled to the client's own hardware, such as NVIDIA DGX Spark, inside an obfuscated virtual machine bound to the TPM 2.0 chip. Bernstein-curve encryption keeps the data and the source inside the perimeter.
Deployment choice
Cloud tenants, dedicated infrastructure or private on-premise patterns are selected according to the product and the organization — not the trend.