Hyperscience has announced a new solution that fine-tunes LLMs with ground truth documents embedded at the core of the enterprise.

Hypercell for GenAI automatically annotates, labels, and structures data from documents for fine-tuning LLMs and GenAI experiences, allowing organizations to rapidly and continuously develop highly accurate, relevant and valuable enterprise models. Through a trusted and proven interface, Hypercell for GenAI promises to accelerate mission critical workflows, grounded in secure, proprietary data, and tuned to the business.

Hyperscience is working with Google Cloud, Hewlett Packard Enterprise (HPE), and other partners on this solution, to give customers flexibility to operate in the infrastructure and AI development platform of their choice, to enable use cases such as prompt engineering, RAG, grounding, and vector search on the customers’ proprietary enterprise data.

"The success or failure of any AI initiative starts with the data that feeds the models," said Andrew Joiner, Chief Executive Officer, Hyperscience.

"Too often, models are built on faulty and incomplete data, and inefficient manual methods and legacy technologies struggle to keep pace with the dynamic flow of documents that course through organizations every day. Today, Hyperscience provides a breakthrough to this challenge by allowing organizations to establish an accurate data estate that trains LLMs to speak the language of their business, and empowers users with relevant, in-context GenAI experiences that align with their business processes and use cases."

Over the past decade, organizations have pursued digital transformation initiatives in order to automate, modernize, and compete in a dynamic and fast-moving marketplace.

Traditional Intelligent Document Processing (IDP), Robotic Process Automation (RPA), and Optical Character Recognition (OCR) technologies have failed to deliver on the promise of digital transformation, since these offerings are rigid and rules-based, and struggle to adapt to new, varied, and complex documents inside an organization.

These solutions have delivered sub-par performance in accuracy, and require significant, expensive manual effort from business process outsourcers (BPOs) to automate processes and workflows inside organizations.

Hyperscience claims to disrupt this paradigm with a novel approach built on AI at its core. Based on a proprietary, machine learning model-based architecture that reads and understands content fluently, Hyperscience says it can deliver accuracy rates of 99.5% and automation rates of 98%.

Hypercell for GenAI establishes a comprehensive data estate to power relevant, in-context GenAI experiences. The solution provides a simple user interface that delivers trusted, accurate results as part of a business user’s workflow.

For example, an insurance claims adjuster could use the Hypercell for GenAI to ask questions in a natural language prompt on the status of a claim. The solution can convert complex documentation such as forms, medical reports, receipts, and doctor’s notes into RAG-ready data for summarization, and provide a recommendation to users on whether to approve or reject the claim based on this ground truth data.

Hypercell for GenAI can run on-premises, in a hybrid cloud, in a public cloud, in a SaaS environment, and even highly secure air-gapped environments. The solution supports a wide range of LLMs, including Mistral Large and Mistral 8X22B, Llama3 (including all three versions), and GPT 3.5 and 3.0.

https://www.hyperscience.com/