Neowise pitches on-device Physical AI for robots and drones

Neowise says its on-device vision, 3D reconstruction and simulation stack can help robots and drones operate without cloud latency or heavy LiDAR dependence. The Pangyo startup is targeting defense, construction and disaster-response markets in the U.S., Japan, the UAE and other global markets.

Why it matters: - Neowise is aiming at markets where robots and drones need to operate continuously, safely and with low latency. - The company is positioning its on-device Physical AI stack as an alternative to cloud-dependent systems and LiDAR-heavy workflows. - The pitch covers defense, border monitoring, construction, disaster response and industrial inspection, all of which depend on fast field decisions.

What happened: - Neowise, led by CEO Charlie Shin, outlined its Physical AI platform in an interview at Startup Campus in Pangyo Techno Valley. - The company says its pipeline combines three products: NeoSight for image restoration, NeoTwin for 2D-to-3D reconstruction and NeoSim for simulation training. - Neowise was founded in April 2024 and is based in Pangyo, Gyeonggi-do, South Korea. - The company has established a Delaware entity to support overseas expansion.

The details: - NeoSight restores images in real time on-device through Neural Processing Units, without separate cloud servers. - NeoSight is designed to improve visibility in dark, low-light, foggy or dusty conditions for robots and drones that need to operate around the clock. - Neowise says NeoSight uses ambient light rather than active infrared, which reduces false detections in military guard and perimeter monitoring settings. - NeoTwin converts standard 2D camera images into 3D digital twins. - Neowise says NeoTwin can cut conversion time from days or weeks, or from 1 to 4 hours with other AI models, to under 10 seconds to about 1 minute. - NeoTwin is intended to replace or reduce dependence on heavy LiDAR sensors that can perform poorly in rain. - NeoSim uses the digital twins to let robots and drones train in simulation before real-world deployment. - The company describes the workflow as Real-to-Sim and then Sim-to-Real. - Neowise says simulated training can reduce costs by more than 95% versus real flight training. - The company says the system is useful in places with no-fly zones or night-flight restrictions because operators can build scenarios from daytime 2D images. - Neowise says its technology can support smart CCTV, border monitoring, Visual SLAM navigation, construction progress checks against CAD blueprints, disaster response and shipbuilding. - In construction and disaster scenarios, the company says NeoTwin can create 3D models quickly enough to support measurement, pathfinding and situational awareness.

Between the lines: - The company is betting that edge processing will matter more than centralized cloud AI in robotics, especially where connectivity, latency and reliability are constraints. - The 3D reconstruction and simulation pitch also suggests Neowise wants to sell into workflows where test conditions are expensive, dangerous or legally restricted. - The Palantir comparison signals an ambition to become a broader operating layer for physical AI, not just a point-solution vendor. - Neowise’s claims are strongest where customers value faster mapping, safer training and lower hardware dependence over traditional sensor stacks.

What’s next: - Neowise says it plans to develop and distribute operating systems for robots and drones. - The company is targeting the U.S., Japan and the Middle East, including Saudi Arabia and the UAE. - Neowise also wants to expand into disaster zones and emergency response use cases, including volcanic, earthquake and structural fire scenarios. - The company says its near-term focus includes construction progress management, safety inspections and hard-to-reach infrastructure such as bridges and power towers. - Neowise previously received TIPS and Early Startup Package support and raised seed funding at an 8 billion won pre-money valuation.

The bottom line: - Neowise is trying to turn robotics AI into a full on-device pipeline, from seeing the environment to simulating the mission to acting in the real world.