Citrusx emerges from stealth to make AI explainable, streamline compliance for enterprises

Tel Aviv-based Citrusx, a startup working to make compliance easier for AI companies and their customers, today emerged from stealth and announced it has raised $4.5 million in a seed round of funding led by Canadian VC Awz, with participation from multiple angel investors.

Founded two years ago, Citrusx is targeting the AI space with a unique focus: making the technology explainable. The company, with its end-to-end platform, allows business stakeholders to understand the AI models they are putting to work, and whether they are working up to the expected mark. 

At a time when countries are exploring the impact of AI and looking to regulate the novel technology, this approach takes AI understanding beyond much-coveted data science teams and experts, ensuring everyone sees what’s going on under the hood and takes the necessary steps to ensure internal and regulatory compliance.

Citrusx: Understanding AI made easy 

Today, AI is a driving force for modern businesses, helping teams accelerate operations, save money and improve customer experiences.

However, when it comes to developing a model and deploying it in production, there are several steps involved — including obtaining input and clearance from non-technical stakeholders in the AI pipeline (such as validation, risk and legal teams).

Now, given the knowledge gap between technical and non-technical teams, the approval part of the process may take a lot of time, delaying the launch of the project. This can easily take 6-9 months, ultimately hitting the business and revenue prospects of the company.

Noa Srebrnik, who previously handled risk management for the Bank of Israel and worked with the country’s ministry of finance, saw the problem well before the gen AI wave hit.

In 2021, she founded Citrusx and started building an end-to-end platform for AI validation and explainability, one that could enable all AI pipeline stakeholders to understand the models in question and perform easy verification and risk mitigation to gain confidence and ensure they’re robust and fair — as per regulatory requirements.

The platform provides clear reports that explain what a model does, how it is performing and how it can achieve business outcomes. Importantly, it covers the entire AI lifecycle, right from development to production. In the development stage, it provides validation metrics for the stability and accuracy of the model, identifies vulnerabilities that could affect its robustness/fairness and provides tools to comply with different regulations. 

Meanwhile, in the production stage, it continuously monitors the deployed model for drifts and prevents its deterioration, explains each new prediction it makes with periodic reporting and suggests steps to reduce potential risks.

“By providing one unified platform for a whole company to use, Citrusx acts as the bridge that connects those mysterious ML models to the rest of the company through validation, monitoring, and reporting. Deployment and maintenance of ML models (can) become smoother and faster, meeting regulatory demands, and helping to explain model decisions to end customers,” Srebnik said in a statement.

In its stealth phase, Citrusx started working with some customers, including publicly listed companies in regulated industries such as finance and security. The names or the number of customers remains undisclosed at this stage but the company claims its patent-pending technology ensures 100% transparency and can lead to 82% faster time to production, enabling teams to quickly seize the AI opportunity. This can give a major boost to teams’ return on AI investments.

“Collaborating across departments and managing multiple deployment models demands seamless communication, and that’s where Citrusx comes in. Its real-time report generation provides high-level insights about models which can be shared between model developers and oversight teams to provide a level of confidence and transparency in the model development process,” a Canadian bank working with the company said.

Much-need offering in the age of ‘AI regulation’

Countries around the world are bullish on the potential of AI, but they are also aware of the risks and are moving aggressively to curb these downsides with regulations. A little over a month ago, President Biden issued an executive order establishing new standards for AI safety and security. And, just recently, the EU finalized a “global first” AI act, a unique legal framework for the development of AI people can trust.

In this phase, tools like Citrusx that help teams navigate regulatory frameworks with robust and fair AI will come in handy. Other players helping teams build better, safer AI are London-based LLM analytics startup Context, which is backed by Google Ventures, and California-based Arize AI, which offers Pheonix, an open-source library to monitor large language models (LLMs) for hallucinations. 

The market is ripe and we expect more players to jump on this bandwagon pretty soon.