I joined Hyperscience to lead and grow the Product Marketing team back in February of this year. What drew me to this market leader in the Intelligent Document Processing (IDP) space, was the combination of AI and Machine Learning deeply embedded at the intersection of apps, platforms, and infrastructure, serving a critical business need. I have been lucky in my career to work at great companies in each of these spaces (Salesforce, Dropbox, and Google Cloud among others), but it’s rare to find a company that combines all three aspects in a unique ML-native solution and go to market strategy.
The IDP landscape is constantly shifting, particularly with the rise of Large Language Models (LLMs). While LLMs are powerful, they aren’t a magic solution for the complexities of enterprise document automation. At Hyperscience, we aim to break the paradigm by providing enterprise-class document automation capable of handling complexity, both visually and linguistically at the highest level. Our approach is rooted in our “Better Together” strategy, enabling seamless interoperability with, and providing essential orchestration, supervision and reporting layers over, advanced AI models like LLMs and Vision Language Models (VLMs).
What I’ve learned as I have come up to speed is the incredible engineering culture and pace of innovation momentum at Hyperscience. We continue to evolve and build out the capabilities of our core AI platform, and this momentum is built upon three key pillars: understanding, speed, and modularity.
Enhanced document understanding for faster decisioning and smarter automation
More powerful models with less risk
Hyperscience includes VLM support and its own next-generation VLM foundation model called Optical Reasoning & Cognition Agent (ORCA), a GPU-optimized model focused on field-level accuracy. With the deployment of VLMs and ORCA, organizations can extend the range of documents that can be automated by handling visually complex and unstructured document formats more effectively.
Full Page Transcription (FPT) is essential for extracting all content from a page. It’s used at a massive scale by organizations like the U.S. Department of Veterans Affairs for processing health benefits claims. FPT is also essential for many generative AI (GenAI) use cases, providing necessary context for Retrieval Augmented Generation (RAG)-based applications, decisioning, or search tasks. The Quality Assurance capabilities of FPT allow organizations to easily monitor and track the accuracy of FPT processing in their document estate.
Knowledge worker post-processing with GenAI
With Hyperscience Document Chat, knowledge workers can interact with documents in post-processing and query them using LLMs within a customized interface, delivering deeper understanding of document context and supporting more informed business decisions. Citations give users the deep context for where the model is extracting specific insights from within each document.
Reduced variability in document processes
Hyperscience enables customers to automatically import and split very long, continuous documents into individual, discrete documents through Auto-Splitting. Users can define rules per layout to split their “box of documents” based on criteria like page length, regex patterns on the first or last page, or an identifier regex (e.g., a specific document number). This eliminates manual document separation, saving significant time and reducing operational overhead.
Business documents are not always uniform; variations in layout, content, or format are common and can lead to processing errors and the need for manual intervention. Traditional systems often require creating new templates for every slight variation, increasing maintenance effort. Hyperscience enhances the Document Drift Management capabilities introduced in R40, and includes a smart, ML-powered capability to detect if an unmatched page group might be a variation of an existing form. This ML-powered approach offers a significant advantage over competitors that might require manual template creation for every minor variation.
Unmatched speed for rapid efficiency, boosting productivity, and lowering costs
Improved accuracy and control at the field level
Enterprises need to ensure the accuracy of extracted data, especially for critical fields that drive key business processes. Hypercell extends the Field-Level Accuracy Targets (FLAT) functionality to Identification models enabling customers to now set accuracy thresholds on specific fields crucial for their business processes.
Automatic error handling & setup of operational tasks in document workflows
Integrating multiple tools for different stages of document processing can lead to complexity, increased costs, and data transfer overhead. Pre-processing steps like file filtering often require disparate software investments. With features such as File Filter Block and Normalization Error sending to Supervision, organizations can now rely on end-to-end capabilities in the Hypercell AI platform to run a more streamlined technology stack.
Enhanced flexibility with a Modular architecture for faster innovation
Infrastructure Enhancements for Scalability and Choice
Enterprises have diverse infrastructure requirements, including preferences for cloud providers and deployment models. The Hypercell, which is available on Google Cloud Platform (GCP), AWS, and Microsoft Azure, provides users with greater flexibility and choice in deployment environments, catering to different enterprise cloud strategies. Additionally, Hypercell deployment on GCP is SOC2 compliant. The latest deployment capability within Google Cloud is Google Distributed Cloud Connected (GDCC), for on-premises deployment of GCP within a company’s own data center.
Modular Architecture
Hyperscience Hypercell continues to build on its modular architecture with composable AI components called Blocks and Flows. The Hypercell activates only the necessary Blocks, optimizing processing time, cost, accuracy, and automation. Workflows can evolve dynamically by incorporating new Blocks and AI models.
Hyperscience Hypercell directly addresses key challenges in enterprise document processing, through advanced AI capabilities, enhanced flexibility, streamlined operations, and a developer-friendly approach. This innovation momentum delivers significant benefits in terms of efficiency, accuracy, cost reduction, and overall automation effectiveness. The modular architecture powers agentic AI applications and positions Hyperscience for future innovation and the ability to tackle increasingly complex document automation scenarios.
Very recently, we spoke with Alan Pelz-Sharpe, founder of Deep Analysis and a leading authority on document automation, about our innovation momentum, and here’s what he said, “There’s a growing realization that while LLMs are powerful, they’re not a silver bullet for enterprise document processing. What sets Hyperscience apart is its ability to orchestrate a mix of models, giving customers the flexibility to choose purpose-built options based on cost, accuracy, efficiency, or time to value. By offering a range of models, from proprietary ML and VLMs to LLMs, Hyperscience customers can choose the right approach based on their business needs, allowing them to build differentiated, proprietary AI strategies that deliver competitive advantage.”
If you’re interested in learning more, we will be hosting a webinar on June 10, titled, “Orchestrating Document Automation with Unprecedented Understanding, Speed, and Modularity”. Click here to register. I’ll be joined by Priya Chakravarthi, Director, Product Management – Application & Machine Learning, Kaloyan Kapralov, Senior Product Manager – SaaS and Platform, and Rich Mautino, Director of Sales Engineering.Look forward to seeing you there!