IN Brief:
- London-based Embedd has raised €2.3 million in pre-seed funding.
- Its platform builds digital models of semiconductor hardware and generates integration artefacts for embedded software.
- The company is targeting faster component evaluation, board bring-up, and software porting across hardware platforms.
Embedd has raised €2.3 million in pre-seed funding to develop software intended to reduce the engineering work required to integrate semiconductor devices into embedded systems.
The London-based company builds digital representations of hardware and uses them to automate parts of board bring-up, driver development, hardware abstraction, and device configuration. Its tooling is aimed at the gap between selecting a component and having reliable application software running on the target hardware.
The funding round was led by Seedcamp, with participation from Cocoa, Connect Ventures, 2100 Ventures, Vesna Capital, U.ventures, Underline Ventures, Common Magic, and Roosh Ventures.
Embedd was founded by Michael Lazarenko, Maxim Gorinov, and Valentin Gololobov and began commercial activity in April 2026. The company says it is already working with semiconductor suppliers, including Microchip Technology, as it develops software support around established embedded ecosystems.
Component integration remains a stubbornly manual part of embedded engineering. A microcontroller, processor, sensor, or communications IC may arrive with hundreds or thousands of pages of documentation covering registers, pins, clocks, interfaces, interrupts, power states, and peripheral behaviour.
That information then has to be translated into board support packages, drivers, device trees, configuration files, and application-specific software. Hardware revisions, operating-system changes, or a switch to a different semiconductor supplier can force engineers to repeat much of the same work.
Embedd’s public MCU Configurator takes a hardware configuration and produces device-tree information for supported targets. Its enterprise tooling goes further by creating a digital model of component behaviour and using that model to generate hardware-abstraction and integration artefacts.
The company’s documentation describes its digital component model as a structured representation of information extracted from a semiconductor datasheet. Project options then adapt the representation to the target design, while deterministic code generation produces the integration output.
That distinction between AI-assisted extraction and deterministic generation is important in embedded development. Engineers can tolerate automation that accelerates repetitive configuration, but low-level software still has to produce predictable register settings, timing, interrupts, memory use, and peripheral behaviour on real hardware.
A plausible-looking driver that invents a register bit or misunderstands a clock dependency is more troublesome than no driver at all. Generated code therefore has to remain traceable to the component model and testable against the physical device.
The same problem grows as embedded systems become more heterogeneous. Robots, vehicles, industrial controllers, medical equipment, and connected devices can combine processors, MCUs, sensors, power-management ICs, wireless devices, and specialised accelerators from several suppliers.
Each additional component brings another software boundary. Even when a device is electrically compatible with the board, its documentation and software package may not fit the operating system, build tools, or abstraction layers already chosen for the product.
Open embedded software platforms such as Zephyr increase pressure for more consistent hardware enablement. Developers want to move between supported devices without rebuilding the software architecture every time, while semiconductor suppliers have an incentive to make their components usable inside those common environments.
Embedd’s tooling is intended to automate enough of the translation work to shorten that path. The company says customers have produced production-ready software for devices up to six times faster using its approach, although the figure is company-reported and will vary according to device complexity and the starting software environment.
The €2.3 million round is modest beside the capital required to develop or manufacture semiconductor silicon, but software enablement has a different cost structure. The immediate spending is on engineering, product development, and vendor integration rather than fabs, masks, or packaging lines.
Embedd now has to show that its models remain accurate as the device catalogue broadens. Automating configuration for one MCU family is useful; reproducing the same reliability across different architectures, operating systems, vendor documentation styles, and rapidly changing silicon revisions is a considerably larger engineering problem.
That is also where semiconductor partnerships become valuable. Direct validation against real devices and vendor data provides a firmer basis for generated integration software than treating public datasheets as infallible machine-readable specifications.
Embedded engineers are unlikely to hand final hardware validation to an automated tool. There is still a substantial amount of repetitive work before that point, however, and Embedd’s funding is aimed squarely at reducing it.
