Johan Brenner 20 January 2025

Backing Sereact: Building a Brain for Robots

The robots are coming! We are on the verge of a profound breakthrough as the worlds of AI and robotics come together. Where robots traditionally struggled to perform tasks in the physical world without specific training, artificial intelligence transforms the way they learn about their environments. The impact of this will be immense. From heavy industry to the home, robots will increasingly replace humans for physical labor, addressing global labor shortages and increasing productivity and growth.

One company at the heart of this is Stuttgart-based Sereact and we are very excited to be backing them in their Series A round. With its large industrial and manufacturing base, Germany has long been a global leader in robotics and has the highest density of industrial robots in Europe. So it’s great to see a German-born company innovating in this space.

Sereact is interesting for a number of reasons. Firstly, it is a pioneer in building Vision Language Action Models (VLAMs) for robotics. This addresses one of the main challenges to date around deployment - that hardware and software had to be specifically designed to perform single, dedicated tasks. Sereact’s VLAM enables zero-shot reasoning where robots can intelligently understand their environment and perform tasks they have not been specifically trained on.

Secondly, Sereact is hardware agnostic. This sets them apart from many companies in the AI robotics space that are hardware-first and many years from mass production (or in some cases currently teleoperated). Sereact’s approach is to build a foundational model that works across all relevant robotics platforms - articulated arms, mobile robots as well as ultimately humanoid. This includes standardized, off-the-shelf robotic hardware. 

Their first product based on this approach, Sereact Pick and Place, automates the pick-and-pack process in warehouses and manufacturing. A key component of this solution is PickGPT, a technical feature launched in 2023 that enables natural language communication with the robot. What makes Sereact Pick and Place particularly valuable is that it can be deployed within a day, delivering immediate cost savings.

Unsurprisingly, this industry-leading time-to-value has struck a cord with customers with the likes of BMW, Daimler Truck, Bol, MS Direct and Active Ants implementing the solution.

In turn, this real world deployment of their models provides valuable data for training. As with any Large Language Model, VLAMs require access to large, diverse datasets to train. Many AI robotics companies rely on synthetic data to do this. Sereact goes a step further with real world data.

Warehousing, logistics and manufacturing is just the start for Sereact. By building the leading foundational model that enables embodied AI for robots they have the potential to become the one of the most important companies in this space. They are undoubtedly one of the best teams in the world for robotics with a proven track record of commercial implementation and we are excited to back Marc, Ralf and the rest of the Sereact team and welcome them to the Creandum family.

Further
articles

Carl Fritjofsson
2025-01-17
Backing nexos.ai: One AI API to Rule Them All
The pioneering AI infrastructure company
Johan Brenner
2024-10-18
Introducing terralayr: The world’s first energy cloud
Enabling energy flexibility on demand
Simon Schmincke
2025-01-14
Boardy: The Future of Networking
The world's most connected AI
Carl Fritjofsson
2024-11-25
Right Here, Right Now. Why we’re backing Appcharge
Appcharge, the leading Direct-to-Consumer platform for mobile game publishers
Carl Fritjofsson
2024-12-18
Kurrent: Making data as available and ubiquitous as electricity
Real-time access to high-quality, reliable, and context-rich data