As artificial intelligence systems become more capable of solving complex problems, the race to develop stronger models is increasingly being shaped by a less visible but critical resource: high quality training data. While much of the AI industry has focused on larger models and greater computing power, many researchers now believe that future breakthroughs will depend on access to expert generated data capable of teaching AI systems advanced reasoning and professional decision making. UK based startup Poindexter Labs is positioning itself at the centre of this shift, building infrastructure designed to help create the next generation of training data for frontier AI models.
The company has secured £2 million in an oversubscribed seed funding round led by Episode 1.
The round also attracted participation from Evertrue Capital, founded by Yvonne Bajela, alongside Octopus Ventures through its First Cheque Fund and a number of angel investors.
Notably, several contributors working on the Poindexter platform, including mathematicians and scientists, also participated in the funding round.
Meeting the Growing Demand for Expert Data
As AI models move beyond simple tasks and increasingly tackle advanced reasoning challenges, demand is growing for specialised training data created by experts.
Traditional data annotation systems were largely designed for straightforward tasks such as image labelling, content categorisation, and text classification.
However, modern AI systems require far more sophisticated inputs.
Fields such as mathematics, science, medicine, law, engineering, and finance demand detailed reasoning processes, nuanced judgement, and domain specific expertise that cannot easily be generated through conventional annotation workflows.
Poindexter Labs aims to address this challenge by building systems that allow subject matter experts to collaborate in generating higher quality training data.
Building a Collaborative Knowledge Platform
The company has translated its methodology into a proprietary platform that is currently operating in beta.
The platform supports both Poindexter’s own data production services and organisations developing advanced AI systems internally.
Through collaborative workflows, experts can create, review, validate, and refine training data in a structured environment designed to improve quality and consistency.
The platform is available to enterprises, frontier AI developers, and public sector organisations seeking better methods for generating expert level datasets.
By focusing on collaboration rather than simple task completion, Poindexter hopes to create a more reliable foundation for advanced AI development.
Rethinking How Training Data Is Created
According to Jocelyn D’Arcy, many existing AI data generation processes prioritise scale and throughput at the expense of knowledge creation.
D’Arcy argues that traditional review systems often encourage contributors to reject work rather than improve it, leading to significant amounts of potentially valuable training data being discarded.
To solve this problem, Poindexter has adopted an approach inspired by academic research.
The company structures its workflows around collaboration, transparency, and peer review, allowing contributors to collectively improve knowledge rather than simply evaluate it.
This methodology is designed to produce richer datasets that better reflect how experts reason through complex problems.
Supporting the Next Generation of AI
The growing sophistication of AI systems is creating demand for increasingly specialised datasets that can teach models how to reason more effectively across professional domains.
Poindexter believes that future AI progress will depend not only on larger models but also on improved methods for capturing and transferring expert knowledge.
Its platform aims to become a key part of that infrastructure by helping organisations generate higher quality training data at scale.
Accelerating Growth
With the newly raised funding, Poindexter Labs plans to expand its team, strengthen relationships with frontier AI laboratories, and broaden adoption of its platform across enterprise and public sector organisations.
As competition among AI developers intensifies, access to expert level training data is emerging as one of the industry’s most valuable assets. By building tools that enable experts to collaborate more effectively, Poindexter Labs is positioning itself to play an important role in shaping the next generation of artificial intelligence systems.