Deep Fusion AI
Category: Autonomous Vehicles
A software-defined 4D imaging radar perception AI that delivers LiDAR-level spatial awareness and object recognition through a proprietary deep learning architecture. Deep Fusion AI was founded in 2022. The company is led by Sunghun Yu. Based in Gyeongsan, South Korea. Team size: 11-50. Total funding raised: $3.0M+. Latest round: Pre-Series A (Jan 2026). Key investors include JNP Global (JNP Adventures), Daegu-Gyeongbuk Technology Holdings (DGTH), Y&Archer, Korea Venture Investment Corp.
- Founded
- 2022
- Headquarters
- Gyeongsan, South Korea
- Team size
- 11-50
- Total funding
- $3.0M+
Value proposition
Provides high-resolution 360-degree perception at a fraction of the cost of LiDAR systems, maintaining high accuracy and reliability in adverse weather conditions like fog, heavy rain, and total darkness.
Products and solutions
RAPA (Real-time Attention-based Pillar Architecture) Perception Software, 360-degree Radar-Vision Sensor Fusion Solution, ECU (Electronic Control Unit) Module Package, Radar-based HD Mapping & Localization Application
Unique value
The 'RAPA' (Real-time Attention-based Pillar Architecture) specifically optimizes radar point cloud processing, achieving over 40% higher detection accuracy compared to existing open-benchmark solutions.
Target customer
Global automotive OEMs, Tier-1 automotive suppliers, autonomous mobility developers, and robotics companies.
Industries served
Autonomous Driving, Robotics, Smart City Infrastructure, Industrial Mobility
Technology advantage
Utilizes a proprietary dataset that incorporates unique radar signal characteristics for training, allowing the AI to distinguish between noise and objects with high precision on edge computing devices without needing expensive hardware.
How they differentiate
Proprietary RAPA (Real-time Attention-based Pillar Architecture) software that achieves LiDAR-level precision using only 4D imaging radar data. Unlike hardware-centric competitors, it focuses on the AI perception layer to achieve 40% higher detection accuracy while minimizing computational load for edge devices.
Main competitors
bitsensing, Smart Radar System, Arbe Robotics, Ambarella (Oculii)
Key partnerships
Global Tier-1 Automotive Suppliers (Ongoing PoC for next-gen sensor stacks), Incheon IFEZ (Smart City & Autonomous Driving testbeds), JNP Global & Daegu-Gyeongbuk Technology Holdings (Investment & R&D partners), TIPS (Tech Incubator Program for Startups) Korea
Notable customers
Chinese Automotive OEM (Guangzhou PoC via Autoroad partnership), Global Tier-1 Automotive Suppliers, Incheon IFEZ (Smart City Testbed)
Major milestones
Founded in Gyeongsan, South Korea in February 2022, Selected for Korean TIPS (Tech Incubator Program for Startup) in April 2023, Secured Seed funding from JNP Global in October 2024, Conducted 4D Imaging Radar PoC with Chinese OEMs in Guangzhou in 2024, Defense Tier-1 contract for unmanned short-range USV system (Dec 2024), Series A Round Open with Pre-Value of $17M USD (June 2025), Defense Tier-1 Unmanned warship system contract (Aug 2025), Won the CES 2026 Best of Innovation Award for RAPA perception technology
Growth metrics
Confirmed Pre-Series A status following significant technical validation at CES 2025/2026; team expanded to 14 experts.
Market positioning
Software-defined radar perception specialist targeting the replacement of expensive LiDAR in autonomous mobility.
Geographic focus
South Korea (domestic R&D), China (PoC with local OEMs), and North America (global automotive Tier-1 partnerships).
Patents and IP
Holds proprietary source technology and multiple patent filings related to radar signal processing, attention-based pillar architectures, and multi-sensor fusion algorithms.
About Sunghun Yu
Sunghun Yu is a veteran in the autonomous driving and artificial intelligence sectors, with specialized expertise in 4D imaging radar-based perception and sensor fusion. Before founding Deep Fusion AI in 2022, he and his core team held senior engineering and leadership roles at prominent autonomous driving technology companies. He has been a pivotal figure in developing 'RAPA' (Real-time Attention-based Pillar Architecture), a software-defined radar technology that aims to replace or augment expensive LiDAR systems in the autonomous mobility space.
Official website: https://www.deep-fusion.com