London-based startup Composo has raised $2 million in pre-seed funding to expand its AI-driven evaluation platform for LLM applications. The funding round, led by Twin Path Ventures with participation from JVH Ventures and EWOR, will support the company’s engineering expansion, client acquisition, and R&D efforts.
Key Points:
- $2M pre-seed funding led by Twin Path Ventures, with JVH Ventures and EWOR participating.
- Composo helps enterprises assess the accuracy and quality of AI-powered apps.
- No-code and API options allow both developers and non-technical users to evaluate AI models.
- Major clients include Accenture, Palantir, and McKinsey, highlighting enterprise demand for better AI monitoring.
Composo aims to solve a growing challenge in AI adoption: evaluating the reliability and quality of applications built on LLMs. Existing methods—such as manual human reviews or using LLMs to assess their own outputs—often lack scalability and consistency. Composo’s proprietary approach combines a reward model trained on human-preferred responses with customizable evaluation criteria for each application.
Unlike competitors like Agenta, Freeplay, and LangSmith, Composo offers a no-code interface in addition to an API, making AI evaluation accessible to a broader range of enterprise users. CEO Sebastian Fox, a former McKinsey consultant, emphasized that the company’s strategy is not capital-intensive. “We’re not building foundation models; we’re making AI more reliable for businesses,” he said.
Composo’s recent launch of its public API for Composo Align, a model designed to evaluate AI applications based on any set of criteria, has already attracted enterprise interest. The startup is focusing on scaling its technology across industries such as healthcare, legal, compliance, and security, where AI reliability is crucial.