Each really sensible machine must understand its environment, however immediately’s AI techniques usually depend on power-hungry {hardware} and fixed cloud connections. That strategy creates a niche between what’s technically potential and what’s sensible for robots, autonomous autos, and different units on the fringe of the Web of Issues (IoT).
Underneath the banner of “Bodily AI,” builders are pushing intelligence out of information facilities and into sensors, cameras, and tiny modules—so machines could make split-second selections with out sending each body again to the cloud. Conventional cloud-based inference can introduce latency, unpredictable prices, and privateness considerations, driving the shift towards on-device AI or Edge AI.
On the coronary heart of this shift lies inference: the section of deep studying the place a educated mannequin takes new inputs—photos, video streams, sensor readings—and delivers real-time predictions. In contrast to coaching, which calls for huge knowledge units and heavy computation, inference is a streamlined “ahead cross” that should occur quick and sometimes with minimal energy.
Edge inference allows a supply drone to detect obstacles on the fly, a manufacturing unit digicam to identify defects immediately, or a subway security system to set off alerts in milliseconds, with out ready on round-trip community delays (latency).
NeoVision, known as the “Mind of Issues” by Neowise, is designed to bridge this hole. Slightly than adapting generic basis fashions, Neowise develops its personal inference engines from the bottom up. By tailoring every neural community and software program stack, NeoVision squeezes most efficiency into tiny footprints. The aim is evident: run subtle imaginative and prescient duties on minimal {hardware} in order that even battery-powered units can function and enormous fleets of vision-capable techniques will be deployed with out onerous infrastructure prices.
To realize its power-saving ambitions, NeoVision faucets two frontier applied sciences. First, Silicon Photonics guarantees as much as 100× higher power effectivity than typical silicon chips by transmitting knowledge as gentle, not electrons. Second, Quantum Computing methods intention to speed up key operations at fractions of immediately’s energy budgets.
Though each fields are nonetheless maturing, Neowise believes this twin-track funding will finally slash inference-chip energy consumption by roughly 35× in comparison with GPU-style {hardware}—and drive complete system prices down by practically two-thirds versus legacy options.
Till lately, techniques leaned on GPUs tethered to cloud servers for heavy lifting. Corporations struggled with 200-watt modules and 350-watt server racks—or confronted the trade-off of lower-precision fashions that missed refined defects or hazards. In the present day’s interim fixes embody hybrid architectures that cut up workloads between the sting and knowledge heart, however they nonetheless add complexity and value.
Against this, NeoVision’s all-in-one chip-and-software answer retains each inference and occasional mannequin updates on the system itself, chopping latency to microseconds and eliminating recurring cloud charges.
Neowise, based on April 12, 2024 is led by Charlie Shin (CEO) alongside co-founders Ned Jo (CTO) and Rocky Son (COO). From its early years, the startup has racked up milestones—an MVP of lively studying by November 2024, a cloud-NPU pilot by December, and partnerships for subway security options in mid-2025. Its staff combines deep experience in AI, photonics, and robotics, backed by a rising patent portfolio and strategic agreements with sensor producers and system integrators.
Neowise CEO Charlie Shin (proper), Ubergizmo Co-founder Hubert Nguyen (left) | Picture by AVING Information
NeoVision’s business technique rests on 4 pillars. First, Neowise will license its proprietary knowledge units and foundational algorithms. Second, it provides vital features—object detection, hazard warning, SLAM—by way of APIs or software program subscriptions. Third, it’s creating its personal NPU chips, cloud platform, and common SDK for seamless deployment.
Lastly, it plans to allow NPU-driven digicam and robotic imaginative and prescient options so companions can construct customized edge-AI techniques with out reinventing the stack. The imaginative and prescient is daring, however as immediately’s Bodily AI wave accelerates, NeoVision goals to change into the go-to “mind” for any system that should see and determine on the pace of sunshine.
Filed in . Learn extra about AI (Artificial Intelligence), GPU (Graphics Processing Unit), Korea, Semiconductors and Startups.
Trending Merchandise
