AIC Lab Research

AI Circuits &
Intelligent Systems

Developing revolutionary hardware architectures for AI — from energy-efficient edge NPU accelerators to neuromorphic computing circuits and in-memory AI processing systems.

Research Gallery

AI Hardware & Memory Interface Design

HBM controllers, high-speed PHY interfaces, and AI accelerator architectures

1/—
AI Accelerator Performance

Key Specifications

Energy-efficient AI computing from edge devices to data centers

100
TOPS · INT8
20
TOPS/W Efficiency
98.5%
Accuracy
5W
TDP
Key Research Topics

Specialized Research Areas

Revolutionizing AI hardware and intelligent systems

🔧

Machine Learning Hardware

Custom AI accelerator architectures optimized for CNN, Transformer, RNN, and LLM workloads. INT8/INT4 quantized inference engines with dynamic precision scaling and 100+ TOPS performance.

📱

Edge AI Computing

Ultra-low-power AI processors for edge devices and IoT. Energy-efficient inference engines, dynamic voltage scaling, and adaptive computing for battery-powered intelligent systems.

🔬

Neuromorphic Circuits

Bio-inspired computing architectures mimicking neural networks in hardware. Spiking neural networks, memristive devices, and event-driven processing for ultra-efficient AI computation.

Applications

Real-World AI Solutions

Breakthrough applications across diverse industries

🤖

Autonomous Vehicles

Real-time object detection, path planning, and sensor fusion for self-driving systems

📷

Spike Neural Network IC

High-performance pattern recognition for surveillance and medical imaging systems

💊

Healthcare AI

Wearable health monitors, diagnostic imaging, and personalized treatment systems

Current Research

Active Projects

Ongoing research pushing the boundaries of AI hardware

🚀
Ultra-Efficient Edge AI Chip
Sub-1W AI accelerator running complex neural networks on battery-powered devices. Dynamic precision scaling, adaptive clocking, and intelligent power gating for maximum energy efficiency.
Silicon Validation
⚙️
In-Memory AI Computing
Compute-in-memory architecture using resistive RAM arrays for neural network weight storage and multiplication. Eliminates data movement overhead for dramatic energy savings.
Prototype Testing
🎯
Adaptive AI Hardware Platform
Reconfigurable AI accelerator dynamically adapting architecture based on neural network topology. Runtime optimization and workload-specific acceleration.
System Integration