Synthgen makes the training images that physical AI can't find in the real world
Synthgen, from Eindhoven, generates labelled synthetic images of rare, safety-critical cases, so computer vision teams can train models that hold up in production.

Synthgen, from Eindhoven, builds the synthetic data layer under physical AI. It generates pixel-perfect labelled training images of the rare, safety-critical cases that are hard or expensive to capture in the real world, so computer vision teams spend less time collecting and labelling data and more time shipping models that hold up in production.
What does Synthgen do?
A camera on a production line, a farm robot or a medical device only works as well as the images its model was trained on. The cases that matter most, such as a rare defect or a dangerous situation, are often the ones that are hardest to photograph. Synthgen generates those missing images, already labelled, and measures the result on held-out real images. Its customers are computer vision teams in manufacturing, agritech, medtech and retail, and the company is expanding into more domains.
"AI for text and AI for robots are completely different problems," says co-founder and CTO Martin Tomov. "There is no internet for robots, so there's no huge pool of data to train large, general-purpose models on." The big physical AI labs now pay people to demonstrate robot movements by hand, one by one, just to record them. General-purpose robots, including robots for the home, still need new research into model architectures. But for many industrial robotics applications, Tomov says, today's models are already more than capable and do not need terabytes of data: "Existing vision models plus a small amount of real data, combined with smart generation of extra representative data, already make a big difference. That's exactly what we do."
Who is behind Synthgen?
Simeon Dimov (CEO) and Martin Tomov (CTO) started Synthgen in late 2024 as a university project. They have since graduated and now work on Synthgen full-time. In 2026 they raised a pre-seed round with OWOW, and Robin Dohmen joined as co-founder. The team is based in Eindhoven, in the Brainport region, and is backed by OWOW, SPARC and Fontys ICT. Synthgen also takes part in the Dutch government's MIT R&D AI collaboration programme (RVO), NVIDIA Inception and Claude for Startups.
What is new?
Synthgen is signing new clients and partnerships across its domains. On 20 October 2026 it hosted "Synthetic Data for Computer Vision & Physical AI" at the AI Innovation Center on High Tech Campus Eindhoven, together with Demcon, an event that drew companies, researchers and students from across the region. Synthgen is a finalist in the AI Pitch Competition, whose finale takes place on 12 November 2026 at the Evoluon. A day later, on 13 November, the founders are the main speakers at "From Fontys to Funded" at Fontys ICT in Eindhoven.
More young AI companies from the Netherlands: new AI startups in the Netherlands in 2026.
Quick facts
- Company: Synthgen
- What it does: generates labelled synthetic training images of rare, safety-critical cases for computer vision and physical AI
- For whom: computer vision teams in manufacturing, agritech, medtech and retail
- Based in: Eindhoven, the Netherlands (Brainport region)
- Founders: Simeon Dimov (CEO), Martin Tomov (CTO) and Robin Dohmen (Co-founder & GTM)
- Started: late 2024, as a university project
- Funding: pre-seed round with OWOW (2026)
- Backed by: OWOW, SPARC and Fontys ICT
- Website: synthgen.co
Sources: Martin Tomov's answers to Meet New Brands (2 November 2026) and synthgen.co.
Learn more at synthgen.co.
Featured in this story? Get your share kit: a ready LinkedIn post and a badge for your website.


