Medly AI raises $8M to give every UK exam student an AI tutor

Every strong company starts with a problem its founders couldn’t ignore. For two London-based doctors turned entrepreneurs, that problem was the widening gap between students who can afford private tutoring and those who cannot. Their solution, an AI-powered exam preparation platform called Medly AI, has just secured $8 million in seed funding to scale its mission across the UK and beyond.

The round was led by Felix Capital, with participation from returning investors Eka Ventures and Ada Ventures. A roster of notable angels also joined, including InvestEngine CEO Andrey Dobrynin, Creator Fund venture partner Al Giles, LearnLaunch general partner Jean Hammond, and Episode 1 general partner Hector Mason. This fresh capital follows a £1.7 million round from February 2025, led by Eka, and arrives alongside a significant public sector win that may hold even more strategic weight.

In June, Medly was selected as one of eight suppliers out of 53 applicants for the AI Tutoring Tools Pioneers Programme, a government initiative run by the Department for Science, Innovation and Technology with the Department for Education serving as delivery partner. Each winner receives £300,000 (excluding VAT) for research contracts running through 31 March 2027, with the goal of boosting attainment among disadvantaged pupils in England. Pearson, ElevenLabs, and Eedi are among the other chosen suppliers.

The founders, Dr Paul Jung and Dr Kavi Samra, have known each other since they were 11, meeting in their first year of secondary school before both went on to study medicine at UCL. Jung, who grew up in Northolt, holds a doctorate in psychiatric neuroscience, while Samra, raised on a council estate in Slough, served as an honorary fellow at UCL Medical School. Neither could afford a private tutor during their school years, a personal limitation that became both the company’s origin story and its core market thesis.

Medly’s product takes a conversational approach to exam practice. Instead of static worksheets, students interact with a system that marks their answers and explains exactly where marks were gained or lost. The platform runs on a proprietary combination of seven large language models, tuned to recognise individual learning styles and adjust feedback accordingly. Handwriting recognition allows students to work with a stylus on a tablet, a feature that proves especially valuable in subjects like maths and chemistry, where showing the working is often as important as the final answer.

The platform currently covers GCSEs, A-levels, and the International Baccalaureate in the UK, and recently expanded to include the SAT in the US. APs and ACTs are slated for release before the end of the year. Looking further ahead, Medly intends to build its own models for UK and EEA marking and pedagogy, rather than relying indefinitely on general-purpose systems.

On outcomes, the company reports that 74% of students who used the platform for last summer’s GCSEs improved by at least one grade, with 41% improving by two or more. These figures come from a survey of Medly’s own users, not an independent assessment. The company also cites subject-specific gains of 46% and 43%, though there is a minor discrepancy in its own materials: the website attributes the 46% figure to English Literature, while the announcement credits Physics.

Growth numbers carry a similar caveat. Medly claims to have surpassed 400,000 users in the UK alone, a figure that sits somewhat uneasily beside the “300,000+ students” currently advertised on its consumer site. According to the company, exam students spend about an hour a day on the platform, exchanging close to a million messages daily.

In June, three team members published a preprint benchmarking 32,534 double-marked GCSE responses across 328 questions and five subjects. The findings suggested that leading off-the-shelf models agreed with examiners at least as closely as two examiners agreed with each other. It is worth noting that the paper tested general-purpose models rather than Medly’s own marking stack, and it has not been peer reviewed. A randomised controlled trial involving more than 1,000 pupils in lower-attaining schools has been completed but remains unpublished.

The social stakes are clear to the founders. “Almost a third of 11 to 16-year-olds receive some form of private tutoring,” Jung said. “For those who can afford it, it’s a powerful advantage. For those who can’t, it’s an equally powerful reinforcer of the status quo.”

Angela Chou, the Felix Capital investor who led the deal, pointed to the founders’ personal narratives and the speed at which they have scaled. Ada Ventures, whose Inclusive Alpha fund backed the pre-seed, has continued its support, as has Eka. Jon Coker, founder of Eka, said the firm had been “blown away by the speed of progress.”

Competition is mounting. Smartschool has been selling a maths reasoning engine into US districts, and a growing number of startups are betting that education-specific models will outperform general chatbots in the classroom. Medly, with its clinical background and government backing, appears well positioned to hold its ground.