Highlights:

  • Citrusˣ’s platform swiftly produces transparent reports detailing an AI model’s functionality, performance, and contribution to achieving desired business outcomes.
  • The Citrusˣ platform delves into every aspect and vulnerability of each AI model, detecting errors and vulnerabilities in its code.

Citrusˣ Ltd., a startup specializing in artificial intelligence validation and explainability, recently debuted with a substantial USD 4.5 million seed funding.

Citrusˣ has developed a user-friendly platform designed for all organizational stakeholders to validate the robustness, explainability, and fairness of their machine learning and AI models. The concept aims to furnish a more user-friendly tool that allows individuals from various backgrounds to validate the precision and dependability of AI. This ensures that data scientists, risk officers, and others can ascertain responsible and fair usage in accordance with regulatory standards.

Utilizing Citrusˣ’s Continuous AI Explainability Platform, organizations can swiftly produce transparent and easily understandable reports detailing the functionality, performance, and alignment with desired business outcomes of an AI model.

The startup contends that such a tool is essential in light of the widespread proliferation of AI-infused applications and systems driven by the popularity of models like ChatGPT and other generative AI models. With increasing numbers of companies rushing to create and implement AI models, the transition from model development to deployment is emerging as a significant bottleneck.

The fact is that only a small number of individuals beyond the data science teams responsible for model development comprehend the inner workings of these models. Consequently, validating, explaining, and assessing the risks of any AI model becomes challenging for individuals, let alone ensuring its compliance with regulations.

Citrusˣ asserts that historically, the majority of companies have relied on generic open-source software to validate their AI models. Nevertheless, these tools are not well-equipped to handle the intricacies of the increasingly prevalent next-generation applications with specialized use cases that numerous organizations are currently developing.

Moreover, they lack accessibility, demanding considerable technical expertise for utilization. Consequently, transitioning a new AI model from prototyping to production deployment can span six to nine months.

The Founder and Chief Executive of Citrusˣ, Noa Srebrnik, highlighted a challenge in AI, emphasizing its complexity. She pointed out that only the data scientists who created the models possess the capability to test and validate them. Consequently, the remaining leadership within a company is left uninformed about the specifics of their AI applications and services. She said, “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.”

According to Srebrnik, this makes deploying and maintaining AI models faster and more seamless. Furthermore, the company’s platform enables organizations to explain their AI model decisions to customers as needed.

The Citrusˣ platform thoroughly examines every aspect and potential weakness in each AI model, detecting errors and vulnerabilities in the code, and highlighting potential bias. It asserts that to address a significant challenge in AI, companies should be provided with assurance of the reliability of the decisions made by their models.

Although Citrusˣ has not disclosed the identities of its clients, it asserts collaboration with several publicly listed companies operating in regulated sectors like finance and security. This includes engagement with a significant Canadian banking institution.

According to Yaron Ashkenazi, the Founder and Managing Partner of Awz, Citrusˣ can significantly impact organizations. He highlights that numerous machine learning models fail to reach production due to apprehensions about their reliability, causing many companies to grapple with obtaining tangible returns on their AI investments. He said, “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.”

Canadian venture capital firm Awz LLC spearheaded the seed round, with involvement from various angel investors, as reported by the company.