top of page

Bespoke AI Hardware

Changing the paradigm of semiconductor design

Our Technology

At Heronic Technologies, we deliver state-of-the-art bespoke accelerator architectures for AI. Inspired by the great mathematician Heron of Alexandria, who pioneered the concept of automata in mechanics, we are redefining the automation of hardware design. We have produced the best-in-class hardware when it comes to TinyML, object detection and human action recognition tasks, setting the industry standard for performance and innovation. Our technology is compatible with the latest FPGA devices from leading vendors, including AMD, Altera, Lattice, Microchip, and GOWIN. Additionally, we are actively developing our ASIC toolflow to further expand our capabilities.

Find out more about the team behind the tech.

54%

MAC Occupancy

Most AI Hardware typically runs at less than 10% for Edge AI applications.

Better Performance over 

Edge GPUs with FPGAs

For object detection applications (read more)

<1ms

Inference Time

Our bespoke architectures massively cut down latency.

AI Model

Specify the model or family of models to accelerate. Supports ONNX, PyTorch and TFLite.

 Technology

Choose the technology platform, whether its a specific FPGA device or an ASIC process node.

Objective

What do you want your hardware to be? Low latency, high throughput? Choose the characteristics of the accelerator

heronIC

AI-Powered Design Automation Toolflow

As we reach the limits of what technology scaling has to offer,  something needs to change in the way we design hardware. At Heronic Technologies, we've realised that the only way to stay ahead in the current hardware landscape is through customisation. That is why we are leveraging cutting-edge AI to automate the design process for the next generation of AI semiconductor devices. We are able to transform our efficient building blocks into world-class accelerator systems.​

Objectively Optimal Hardware

Rapid Design Cycle
Extremely High Efficiency
Microsecond Latency
Best Performance Per Dollar

Applications

Using our technology, we are able to create AI accelerator architectures for a number of different applications, both for ASIC designs and FPGA devices (Altera, AMD, Lattice).

Object Detection

cars_edited.jpg

Responsive, low latency inference to provide instant actionable intelligence.

Low-latency object detection is essential for applications where milliseconds make the difference. This technology drives real-time quality control by instantly flagging defects on high-speed industrial lines and is fundamental to safety in automotive and aerospace, enabling vehicles to immediately react to hazards and ensuring precision in complex assembly.

Super Resolution

football_pixelated_edited.jpg

Enabling cost-effective, high-quality video upscaling.

Traditional upscaling techniques have run their course in today's world of ultra-high resolution videos. We have been developing IP that can deliver AI-powered super resolution, transforming low resolution video sources all the way to 8K and beyond. This can provide unparalleled image quality for medical devices, broadcast and defence applications.

Image Segmentation

cells_edited.jpg

Real-time segmentation directly from the sensor.

Modern image segmentation demands immense computational power to classify every pixel in an image, often in real-time. Our hardware accelerator designs are crucial for overcoming these challenges at the edge, enabling the deployment of complex neural networks that can perform precise, instantaneous segmentation for applications like robotic surgery and autonomous driving, all while operating within strict power and thermal limits.

Our News

Find the latest articles on Heronic below. You can also sign up to our newsletter.

Heronic @ Future of Computing Conference 2025 London
26 Mar 2025
Heronic joins DBT led Deep Tech Trade Mission to Seoul
18 Feb 2025
Heronic won Rohm Semiconductor's reverse-pitch competition
3 Oct 2024

Contact

Get in Touch

Have a question or want to learn more about our technology? Reach out to us today.

bottom of page