Big-data readiness startup Illumex Technologies Inc., which aims to overcome the challenges associated with structured information, said today it has closed on a $13 million seed funding round.

Today’s round was led by Cardumen Capital, Amdocs Ventures and Samsung Ventures, and saw the participation of ICI Fund, Jibe Ventures, Iron Nation Fund, Ginnosar Ventures, ICON Fund, Today Ventures and a host of angel investors.

Though it’s well-known that it can be challenging for generative AI models to tap into the value held within unstructured data, such as images, videos, audio files and handwritten notes, far fewer people are aware of the headaches involved in making structured information accessible to those models.

It’s these headaches that Illumex is trying to overcome by automating the creation of a semantic layer that unifies structured data silos, adding business context and generating a consistent library of domain-specific terminology to make structured information easier for large language models to read.

Structured data struggles

Illumex believes hundreds of organizations might be interested in its “Generative Semantic Fabric,” for enterprise adoption of generative AI has grown exponentially over the last year. According to a recent study by McKinsey & Co., more than two-thirds of large enterprises have already adopted some kind of generative AI technology, and it’s already driving significant value for those businesses. Despite this, concerns remain about the accuracy of generative AI and the risks of AI “hallucinations,” where AI models generate inaccurate responses to user’s queries and prompts.

Some companies are looking to solve the hallucination problem by integrating retrieval-augmented generation, or RAG, which is a process that allows them to make their own, proprietary data available to their generative AI models. The idea is simple enough: Enhance the underlying large language models’ knowledge base by giving it access to more data, specific to the business operations. But doing this is not easy, and a huge number of companies have reported drawbacks with RAG, especially when it comes to structured data.

One of the chief problems with RAG pipelines is that they require continuous manual maintenance, which can take up a huge amount of time and effort. The other is that RAG is unable to provide a single source, given the lack of any semantic meaning in structured information.

Although unstructured data is semantically meaningful, structured information held in traditional databases lacks a lot of context. It needs to be manually labeled, or else there will be a lot of ambiguity and inconsistency in the outputs of the generative AI models it’s fed into.

Adding semantic meaning and context

Illumex aims to remedy these headaches by using generative AI itself to automatically discover, map and add semantic meaning to structured data. It does this by analyzing the metadata associated with structured datasets, without accessing the actual information within, in order to create a unified semantic knowledge graph that acts as a single source of truth, aligning any data that’s inconsistent with business terminology and context.

The company’s service automatically builds what it says is a “domain-specific ontology” that formally defines all of the entities, properties and relationships within companies’ knowledge structures, enabling more consistent and contextually-relevant data interactions. This ontology can then serve as the foundation for LLMs to map users’ questions and prompts reliably and accurately to the relevant information it should retrieve in order to provide a proper response, the startup says.

Founder and Chief Executive Inna Tokarev Sela said the success of generative AI projects depends on solving the fundamental challenges around structured data.

“Our Generative Semantic Fabric aligns organizational data with business meaning and domain-specific context, allowing organizations to finally trust the results of their AI initiatives,” she said. “By enabling business users to interact reliably with data using natural language, without teaching them the precise technical definitions, we’re democratizing AI and empowering enterprises to make better decisions.”

Illumex had been operating in stealth mode until now, but it has already signed up a number of noteworthy enterprise customers, including Teva Pharmaceutical Industries Ltd. and Carson Optical Inc. In addition, it claims to have forged partnerships with the big three cloud infrastructure providers – Amazon Web Services Inc., Google Cloud and Microsoft Corp.

Cardumen Capital’s Gonzalo Martinez de Azagra says that generative AI is an existential imperative for many organizations, especially those in more complex and highly regulated industries.

“They face the unique challenge of needing to balance AI-powered innovation with strict data governance and security requirements,” he added. “Illumex perfectly bridges this gap, enabling enterprises to maximize the generative AI opportunity while ensuring the integrity, lineage and trustworthiness of the data fueling these applications.”

Image: SiliconANGLE/Microsoft Designer