Hyperscience announces the Spring 2026 Hypercell release, expanding its IDP-focused infrastructure to enable intelligent inference across diverse hardware and models for higher accuracy at lower cost.

The platform now emphasizes production-grade security and scalability, including FedRAMP High authorization, advanced redaction with synthetic data for PII, high-throughput AI-in-the-loop governance, and auto-scaling across SaaS, managed clouds, and air-gapped environments.

Real-world adoption is highlighted with a major US mortgage provider using Hypercell for income and asset verification via ORCA VLM, plus a global audit and tax firm automating tax and invoice workflows with Gemini under governance and cost-control measures.

A key framework update introduces an open, model-agnostic VLM framework that supports latest hardware and models, including NVIDIA Blackwell GPUs, NVIDIA Nemotron, and Google Gemini 1.5/2.5, with built‑in thresholding, fine-tuning, QA, observability, and orchestration of VLMs and frontier models.

Contacts for media inquiries: Jyotsna Grover.

ORCA, Hyperscience’s VLM, gains enhancements like one-click deployment, pre-built subflows, training options to improve accuracy, and new model management tools for business users to monitor and iterate model performance while preserving governance and data security.

Hyperscience presents as a market leader in enterprise AI infrastructure for IDP, with clients including American Express, HMRC, Mars, Stryker, U.S. SSA, and U.S. VA, backed by investors Bessemer, Battery, FirstMark, Stripes, and Tiger Global.

A core feature is inference-layering optimization through a unified orchestration layer that dynamically balances workloads across CPUs, GPUs, and various model architectures to maximize accuracy and minimize cost per transaction.

Hyperscience will showcase Hypercell innovations at Google Cloud Next in Las Vegas (April 22–24) and host a Spring 2026 release webinar on May 14; CEO Andrew Joiner will discuss the “Inference Inflection Point” for trusted data pipelines.

The Spring 2026 release advances a Kubernetes-for-IDP approach, enabling developers to build trusted data pipelines with cutting-edge models and hardware in a production-grade, secure environment.