Artificial intelligence validation and explainability startup Citrusˣ Ltd. is launching today armed with $4.5 million in seed funding.

Citrusˣ is the creator of what it says is an easy-to-use platform that will allow all stakeholders within an organization to check that their machine learning and AI models are robust, explainable and fair. The idea is to provide a more accessible tool that anyone can use to verify the accuracy and reliability of AI, so data scientists, risk officers and others can ensure it is being used responsibly and fairly, and in line with regulatory requirements.

With Citrusˣ’s Continuous AI Explainability Platform, companies can quickly generate clear and accessible reports that explain what an AI model does, how well it performs, and whether or not it is helping to achieve the desired business outcomes.

The startup argues that such a tool is necessary given the rampant growth of AI-infused applications and systems that was sparked by the popularity of ChatGPT and other generative AI models. As more companies scramble to build and deploy AI models, the journey from model development to deployment is becoming a key bottleneck.

That’s because very few people outside the data science teams that develop models understand how they actually work. As a result, it’s difficult for anyone else to validate, explain and assess the risks posed by any AI model, let alone ensure it is compliant with regulations.

Citrusˣ says that traditionally, most companies use generic open-source software to validate their AI models. However, these tools are not well-suited to the more complex, next-generation applications with specialized use cases that many organizations are building today.

They’re also not very accessible, as significant technical expertise is required to use them. As a result, it can take anywhere from six to nine months for a new AI model to go from prototyping to production deployment.

Citrusˣ founder and Chief Executive Noa Srebrnik said the problem with AI is that because it’s so complex, the only people who are able to test and validate models are the data scientists who designed them. That means the rest of the company’s leadership is left in the dark about the nature of their AI applications and services. “By providing one unified platform for the whole company to use, Citrusˣ acts as the bridge that connects those mysterious machine learning models to the rest of the company through validation, monitoring and reporting,” she said.

The result is that the deployment and maintenance of AI models becomes a much smoother and faster process, Srebrnik insisted. The company’s platform also allows organizations to explain their AI model decisions to customers, when necessary, he added.

Citrusˣ’s platform digs down into all of the elements and weak spots in each AI model, identifying any errors it makes and any vulnerabilities in its code, while highlighting potential bias. It claims to be solving one of the major pain points of AI, helping companies to know that they can trust the decisions their models make.

Citrusˣ hasn’t named any of its customers, but claims to be working with a number of publicly listed companies in regulated industries such as finance and security, including a “major Canadian bank.”

Awz founder and Managing Partner Yaron Ashkenazi said Citrusˣ can make a real difference for organizations. He points out that many machine learning models never make it into production due to concerns over their reliability, and so many companies are struggling to obtain any real return on investment in AI. “By helping the whole organization to understand what an ML model is doing, and limiting risk, Citrusˣ is instilling the confidence required to go live, while ensuring optimal company performance,” he said.

The seed round was led by Canadian venture capital firm Awz LLC, and saw participation of several angel investors, the company said.

To read the full story on SiliconANGLE click here.