Paris, France-based AI infrastructure startup Weflayr has raised $2.5 million in Pre-Seed funding to develop its AI inference routing platform, which helps companies select and manage AI models based on their performance on real workloads. The round was led by 13books Capital, with participation from Founders Future, Kima Ventures, Plug and Play, Station F Ventures, Roxanne Varza, a General Catalyst scout, and several unnamed angel investors. The company will use the funding to expand its engineering and AI research teams, accelerate product development and reach more AI companies across Europe and the United States.
Founded by Romain Mazeries, CEO and former CEO of Mangopay, and Noé Achache, CTO and former Head of AI Engineering at Theodo Data & AI, Weflayr is developing infrastructure designed to help AI companies manage the trade-offs between model quality, performance and inference costs. Its platform evaluates models using customers’ own production workloads rather than relying exclusively on general-purpose benchmarks, allowing businesses to define which models can handle different requests.
Optimising AI Model Selection Through Intelligent Routing
As the number of available AI models continues to grow, companies building AI-powered products face increasingly complex decisions about which models to use. Different models offer varying levels of performance, response quality and cost depending on the task, while new releases can quickly change those trade-offs. Traditional AI gateways allow businesses to distribute requests between providers, but routing decisions may not reflect how individual models perform on a company’s specific workloads.
Weflayr addresses this challenge by combining model benchmarking, dynamic routing and cost analysis in a single platform. Companies can test different AI models against their own production data, evaluate results according to quality and cost requirements, and approve the models they want to use. When a new request arrives, Weflayr matches it with similar previously benchmarked calls and routes it to the best-performing approved model within the customer’s defined parameters.
The platform also connects inference decisions with business metrics, allowing companies to monitor costs and margins across customers, product features and workflows. Through a unified AI gateway, developers can access multiple model providers while managing routing rules, fallback mechanisms and usage limits. This approach is designed to give businesses greater control over their AI infrastructure without requiring them to select a single model for every task.
According to Weflayr, its first deployment demonstrated a reduction of more than 55% in inference costs while maintaining equivalent response quality. The company also reported a 15-percentage-point improvement in relative quality at equivalent cost compared with a frontier model. These results come from the company’s initial deployment and have not been presented as independently verified benchmarks across multiple customers.
Funding to Expand Engineering and AI Research
Weflayr was founded by two executives with experience in financial technology infrastructure and production AI systems. Romain Mazeries previously served as CEO of payments company Mangopay, while Noé Achache led AI engineering at Theodo Data & AI. The founders are applying their experience in infrastructure and automated decision-making to AI inference, where companies must continuously balance model performance, reliability and operating costs.
The $2.5 million investment will support the expansion of Weflayr’s engineering and AI research teams, further development of its routing technology and commercial growth across Europe and the United States. The company is currently onboarding its first customers through a private beta programme and preparing for broader product availability in the coming weeks.
Weflayr is also recruiting engineers and technical specialists in Paris as it develops its infrastructure for production AI applications. Its longer-term objective is to provide a system that continuously evaluates model performance and adjusts routing decisions according to each customer’s requirements, helping AI companies manage inference costs while maintaining control over quality.
With the new funding, Weflayr plans to move beyond its initial deployments and expand access to its platform as businesses increasingly adopt multiple AI models across their products and services.
