Hypercell Spring 2026 advances Inference Layering Optimization, enabling enterprises to dynamically route workloads across latest CPUs, GPUs, specialized and frontier models to achieve mission-critical accuracy at the lowest cost per transaction

NEW YORK: Hyperscience, a market leader in enterprise AI infrastructure software, focused on Intelligent Document Processing (IDP), today announced major advancements to its Hyperscience Hypercell platform as part of the Spring 2026 release.

Innovations include:

• Inference layering optimization – a unified orchestration layer that dynamically and intelligently balances workloads across CPUs, GPUs, and diverse model architectures to deliver the most accurate and affordable automation.
• Open, model-agnostic VLM framework innovation – an open and agnostic approach that enables developers to leverage the latest AI hardware and models, including support for NVIDIA Blackwell GPUs, NVIDIA Nemotron 3, and Google Gemini 1.5 Flash and Gemini 2.5 Pro. Going beyond simple model wrapping, Hyperscience makes models first-class citizens within the enterprise inference stack through built-in thresholding, VLM fine-tuning, QA, observability, and orchestration that directs VLM and frontier models to the highest-value, most complex inference tasks. Hyperscience is also the only company in the market with an out-of-the-box, accuracy-harnessed VLM framework that empowers knowledge workers to instantly deploy subflows, train models, and manage advanced prompts for day-one automation of complex enterprise documents.
• Production grade by design – provides industrial strength infrastructure required to support document centric AI initiatives at scale, including auto scaling, FedRAMP High authorizations, advanced redaction and masking with synthetic data for stringent PII handling, compliance, and high throughput AI-in-the-loop governance and performance.

“The enterprise is moving past the experimental phase of GenAI and into a new era of industrial-scale, mission critical production use," said Andrew Joiner, CEO of Hyperscience. "The Hypercell Spring 2026 release represents an inference inflection point. By providing a unified orchestration layer that intelligently balances accuracy with cost, we are giving organizations the trusted data foundation and high-performance infrastructure they need to transition from simple automation to a truly agentic enterprise.”

Inference layering optimization delivers accuracy, automation, and affordability

Chief technology officers, developers, and business leaders today face a bewildering array of choices in pursuing an effective AI strategy that delivers accuracy, automation, and affordability. However, getting there is not easy. Throwing every document-processing inference task to expensive GPU-based models will quickly put organizations over budget. Conversely, relying on legacy tools and manual effort from business process outsourcing (BPO) will result in lower rates of automation and accuracy.

These challenges are compounded when considering the essential role that documents play in fueling both systems of record and systems of AI. Without a trusted platform for delivering ground truth data into mission-critical business workflows as well as the AI factories that create agentic experiences – both systems grind to a halt.

Hyperscience uniquely addresses these challenges through an intelligent orchestration architecture that seamlessly directs document processing tasks to the most appropriate model, based on the particular demands of each workload and use case. Utilizing intelligent inference layering, the platform routes high-volume, straightforward transactions to efficient CPU-based models to minimize costs. When workflows require deep context or complex reasoning, the Hypercell dynamically shifts to a GPU-centric layer - deploying heavy-duty Vision Language Models (VLMs) such as the Hyperscience Optical Reasoning and Cognition Agent (ORCA) to perform the sophisticated calculations and policy based recommendations necessary for mission-critical production.

This unique approach builds on deep ML model development, management, and orchestration expertise inside Hyperscience, and is proven in the real world with key design-partner customers.

For example, one of the largest mortgage providers in the United States relies on Hyperscience for income and asset verification. The company leverages Hypercell’s inference layering optimization to apply a mix of models – spanning specialized models as well as the ORCA VLM – to achieve best-in-class understanding and accuracy of mortgage applicants’ bank statements and paystubs. This enables faster decisioning while maintaining the accuracy and affordability required for mission-critical lending operations.

Likewise, a global audit, tax, and professional services firm is leveraging Hyperscience to orchestrate complex tax and invoice workflows, combining Hypercell models with Google Gemini 1.5 Flash and 2.5 Pro. By applying intelligent orchestration and model thresholding, this customer is scaling automation in complex audit scenarios, while maintaining strict control, governance, and cost management.

Open model-agnostic VLM framework innovation delivers superior accuracy and accelerates time-to-value

An inference layering strategy is only as valuable as the depth and variety of models and technologies available to an organization. Hyperscience continues to add support for the latest AI hardware and models, ensuring that customers can leverage the broad AI ecosystem as first-class citizens inside the Hypercell platform.

The platform puts the power in the hands of developers – including a native GenAI model library – that allows them to build document processing pipelines that leverage the best of the industry ecosystem.

The Spring 2026 builds on this “Kubernetes for IDP” concept by allowing developers to build trusted data pipelines with cutting edge models and hardware in the AI ecosystem, in a safe, production-grade environment:

• NVIDIA Blackwell GPUs: To process complex tasks rapidly and efficiently, the Hypercell platform fully supports NVIDIA Blackwell GPUs, showcasing a commitment to enabling customers with cutting-edge infrastructure.
• NVIDIA Nemotron models: Hypercell now supports seamless orchestration of GPU-accelerated models for complex reasoning and agentic automation.
• Google Gemini (1.5 Flash & 2.5 Pro): Hypercell includes native support for advanced tasks like summarization, classification, and content reasoning within orchestrated workflows.

To accelerate time-to-value and eliminate traditional IT bottlenecks, the Hypercell Spring 2026 release also includes updates to the ORCA framework that give business users unprecedented, out-of-the-box control over their document automation workflows:

• One-click ORCA deployment and out-of-the-box subflow: With this release, Hyperscience has made it even easier to see immediate value with the ORCA VLM Framework. Organizations can instantly install ORCA and a pre-built subflow that provides everything needed to test the model on production data. This reduces onboarding friction and accelerates time-to-value.
• Increased accuracy and automation with ORCA training option: While ORCA's VLM capabilities allow clients to recognize significant value without any training, customers can now opt to train ORCA models on their documents within the application through a user-friendly Training Data Management interface. This provides customers an option to further increase accuracy and automation, reduce time to value, and eliminate data sharing risks.
• Improved model management for business users: With ORCA’s new model management tools, business users gain unprecedented control over how the system operates. This ensures that outputs are consistently reliable and tailored to each organization’s unique requirements. Additionally, the platform provides clear monitoring and versioning features, allowing teams to track model performance, make iterative improvements, and deploy updates safely, all without involving IT teams or compromising data security.

Production-grade by design platform provides secure, compliant, and scalable document automation:

Hyperscience brings all this innovation – inference layer optimization, access to a rich AI ecosystem, and a VLM framework – onto the Hypercell, a secure, auditable, and scalable enterprise-class platform. This makes the Hypercell the ideal platform for customers in regulated industries – as well as any organization seeking a robust platform for production-grade document automation.

• Secure and compliant: the Hypercell delivers best-in-class security standards, including FedRAMP High authorization, and provides unmatched redaction, masking, and synthetic data replacement capabilities for organizations seeking train models in a safe, compliant manner.
• Versatile and scalable: Powered by ORCA, Hypercell provides native agentic AI for IDP. These autonomous agents reason through document structure and intent to validate, route, or escalate data without custom code. This self-optimizing orchestration increases throughput and accuracy, integrating seamlessly with enterprise workflows while maintaining full explainability and human-in-the-loop control.
• Delivered across all environments: the Hypercell is a fully containerized architecture that can be deployed across all infrastructure models – including fully managed SaaS, customer managed private and public clouds (including AWS, Microsoft Azure, and Google Cloud), as well as air-gapped environments.

Engage with Hyperscience in person and online

• Hyperscience will showcase these innovations at Google Cloud Next from April 22-24, in Las Vegas, Nevada, at booth #1907 where attendees can see the Hypercell in action and learn how to operationalize AI at scale.
• Hyperscience will host a Hypercell Spring 2026 release webinar on May 14, offering a deeper dive into the platform’s latest capabilities and real-world applications.
• Read the blog from Hyperscience CEO Andrew Joiner: The Inference Inflection Point: Building Trusted Data Pipelines for the Agentic Enterprise.

About Hyperscience

Hyperscience is a market leader in enterprise AI infrastructure software, focused on Intelligent Document Processing (IDP). The Hyperscience Hypercell platform unlocks the value of an organization’s back office data through the automation of end-to-end processes, and transforms complex documents into LLM and RAG-ready data to power new enterprise GenAI experiences. This enables organizations to transform manual, siloed processes into a strategic advantage, resulting in a faster path to decisions, actions, and revenue; positive and engaging customer, public, and patient experiences; and dramatic increases in productivity.

Leading organizations across the globe rely on Hyperscience to drive their hyperautomation initiatives, including American Express, Charles Schwab, HM Revenue and Customs, Mars, Stryker, The United States Social Security Administration, and The United States Department of Veterans Affairs. The company is funded by top tier investors including Bessemer Venture Partners, Battery, FirstMark, Stripes, and Tiger Global.

Fonte: Business Wire