Monday, 16 November 2026
Rīga TV

World and Latvian news in one place

TechnologyPublished: 16 November 2026 at 08:01

German startup Daisytuner raises €6 million to speed up AI models across different chips

Darmstadt-based startup Daisytuner, which automates AI model performance engineering for any chip, has raised €6 million in Seed funding. It also launched Daisyflow, an inference engine.

Foto: EU-Startups

Daisytuner, a startup from Darmstadt, Germany, that automates the performance engineering of artificial intelligence (AI) models on chips from any vendor, has raised €6 million in Seed funding. The issue it targets is that computing capacity is scarce in the AI industry and much of it is tied to a single vendor's software. Alongside the round, the company launched Daisyflow, an inference engine built to run AI models efficiently on any hardware.

The round was co-led by European venture capital firms Elaia and LEA Ventures. Amadeus APEX Technology Fund and Seedcamp also took part.

How it works

Daisytuner was founded in 2024 by Lukas Trümper and Adrian Schmitz. Its team draws on more than a decade of research at institutions including ETH Zurich, TU Munich, RWTH Aachen, TU Darmstadt and Lawrence Livermore National Laboratory. The company combines a self-learning compiler with a patented optimisation platform.

Its open-source compiler, DOCC, converts code from popular frameworks such as PyTorch into native code for each target chip, including NVIDIA and AMD GPUs and newer architectures. AI agents test optimisations on real hardware around the clock, and each verified result is stored in a performance database and reused for similar code.

This work, known as performance engineering, has historically been manual and done mostly by a handful of leading labs. As hardware grows more varied, each new chip, GPU generation or model can break earlier optimisations.

Daisyflow and plans

Daisyflow serves language, vision, speech and recommendation models through APIs that engineering teams already use, and it can be deployed inside a customer's own infrastructure. In a company statement, Daisytuner said the platform ran Qwen 2.5 up to 18% faster than the industry-standard vLLM on AMD Instinct GPUs. It is available through the company's Cloud Console.

The startup intends to use the funds to expand Daisyflow's reach among enterprises, grow its engineering team and widen compiler support for emerging chip architectures. Sébastien Lefebvre, Partner at Elaia, said the founders have built high-level compiler technology. Trümper, co-founder and CEO, said the company is working with chip vendors, European hyperscalers and enterprises.

Comments

0/1500

Comments are automatically moderated. No hate, threats, personal data or spam.

Loading comments…

More in this category