Humanloop: helping organisations deploy AI faster by putting humans in control

Problem to be solved

Machine learning can deliver major benefits, but building high-performing models often depends on large volumes of high-quality labelled data. 

For many organisations, data annotation becomes a bottleneck-time-consuming, expensive and reliant on scarce subject-matter experts. Teams need ways to train and iterate models faster, with less manual labelling, while maintaining quality and accountability.

Solution

Founded by UCL Professors David Barber and Emine Yilmaz with PhD students Raza Habib and Peter Hayes, Humanloop developed tools that help teams build language and other ML systems more efficiently by keeping experts ‘in the loop’ only when they add the most value. By using probabilistic approaches that capture uncertainty, the system prioritises the examples that will improve a model fastest, reducing the volume of labelling needed and supporting quicker deployment and iteration.

How UCL Ventures helped

Humanloop was spun out with support from UCL Ventures through the Portico Ventures model, which provides a straightforward route for software and know-how based businesses to form and license relevant IP from UCL. UCL Ventures supported early-stage set-up and investor conversations, and the company was eligible to apply for investment from the UCL Technology Fund (UCLTF), alongside wider connections across UCL’s innovation ecosystem.

As Raza explains: “The range of applications for natural language understanding has grown rapidly in the last couple of years,”

Where is Humanloop now?

Humanloop launched an early product in September 2020 and was accepted into Y Combinator the same year, helping sharpen its offering and build momentum. The team has worked with customers on applications such as content moderation, document and contract classification and monitoring tasks, and continues to refine its platform so that more organisations can deploy reliable AI systems with less manual effort. 

Next, the focus is on scaling adoption and extending capabilities to support new model types and use cases. In 2026, Humanloop joined AI company Anthropic, its statement reads: "Humanloop team is joining Anthropic!

From the start, our mission at Humanloop has been to enable the safe and rapid adoption of AI. Now, as the pace of AI progress accelerates, we think Anthropic is the ideal home to amplify our impact.

We've had the privilege of working with customers who continually push the boundaries of AI adoption. We owe a huge debt of gratitude for their trust and invaluable feedback along the way. As we sunset the Humanloop platform, we will continue to work closely with our customers to make their transition as smooth as possible."