Today's AI chips compute even when nothing happens
Every clock tick wakes thousands of compute units at once. Even when the input has not changed at all, the power is spent all the same.
Neuromorphic AI Processor
Compute like a brain.
MindCore computes only at the instant a spike arrives. Event-driven instead of clock-driven, it runs real-time AI on a battery.
FPGA prototype verified · ASIC in development
100×
Energy efficiency (up to)
<1ms
Inference latency
1M
Neuron scale
<1W
Target power
The shift
MindCore wakes only where a spike lands
Like a biological neuron, computation happens only where something changed. The rest of the fabric stays silent, and silence costs nothing.
Same input, same decision, a fraction of the work
Event-driven processing combined with temporal coding cuts energy by 10-100x while holding accuracy.
Clock-driven
100%
Event-driven
—
Product
MindCore
A spike-driven, programmable neuromorphic accelerator. The efficiency of a biological brain, cast into silicon for on-device AI.
- Architecture Spike-driven programmable core
- Neuron model Hardware LIF neurons
- Learning STDP-based on-chip learning
- Scale-out Multi-chip interconnect
Biological Neural Modeling
Leaky Integrate-and-Fire neurons implemented directly in hardware, reproducing the behaviour of real neural circuits.
Event-Driven Processing
An asynchronous fabric that computes only on spike events, eliminating wasted switching power.
Real-Time Learning
On-chip STDP learning brings real-time adaptation to the edge.
Scalable Architecture
A multi-chip scalable architecture for building large neural systems.
Low Latency Inference
Sub-millisecond inference for applications that must decide immediately.
Energy Harvesting Ready
Ultra-low power design that can run on harvested solar or vibration energy.
Where we are
- Architecture research Published at MCSoC 2025
- FPGA prototype Verified on hardware
- ASIC implementation In progress
- Commercial release Full specs to be announced
Technology
Four pillars
From algorithm to silicon to deployment, we design the entire neuromorphic stack ourselves.
Spiking Neural Networks
Network architecture modelled on biological spiking neurons, native to temporal information.
Event-Driven Computing
Computation is triggered by events, removing redundant work and power at the source.
Deep Learning Optimization
Training and quantization techniques that lift SNN accuracy to conventional deep-learning levels.
Edge AI Solutions
An ultra-low-power on-device AI stack for mobile and IoT hardware.
Demo
Running on real hardware
Not a simulation. This is MindCore inferring in real time on the FPGA prototype board.
Applications
Anywhere power is precious
Computer Vision
Real-time object recognition and tracking with event cameras
Medical AI
AI diagnostics for low-power wearable medical devices
Autonomous Driving
Real-time perception and decision making
Robotics
Biologically inspired sensorimotor control
IoT & Smart Sensors
Ultra-low-power edge intelligence
Brain-Computer Interface
Neural signal processing and interpretation
Research
From paper to silicon
Our technology stands on peer-reviewed research.
2025
MindCore: Spike-Driven Programmable Accelerator for On-device Neuromorphic Computing
IEEE International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC 2025)
@inproceedings{park2025mindcore,
title = {MindCore: Spike-Driven Programmable Accelerator for On-device Neuromorphic Computing},
author = {Park, Hawon and Lee, Si Yong and Lee, Ryangjin and Kim, Yoora and Yang, Yoon-Seok},
booktitle = {IEEE International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC)},
year = {2025}
}
Ultra-low-power AI
Power efficiency from event-driven processing
High-performance learning
Accuracy on par with conventional deep learning
Biological inspiration
Design that follows how the brain actually works
Real-time processing
Fast inference that exploits timing
Team
The people building it
Engineers who have lived in both worlds: neuroscience and silicon.
Prof. Yoon-Seok Yang
Founder & CEO
Assistant Professor, Department of Computer Science, SUNY Korea
Assistant professor in Computer Science at SUNY Korea. Previously a Tensor Processing Unit (TPU) silicon and research engineer at Google in Sunnyvale, California. Before Google, he was a research scientist at the Neuromorphic Computing Lab at Intel Labs in Santa Clara from 2012 to 2022, working on neuromorphic computing systems and AI chip design. He earned his Ph.D. in electrical and computer engineering from Texas A&M University.
- Ex-Google TPU
- Ex-Intel Neuromorphic Lab
- Ph.D. ECE, Texas A&M
Research Team
SY
Si Yong Lee
Principal Research Engineer
HP
Hawon Park
Senior Research Engineer
RL
Ryangjin Lee
Research Engineer
Contact
Let's build it together
Technical collaboration, pilot deployments and investment enquiries are all welcome.
yoonseok.yang@sunykorea.ac.krIncheon, South Korea