Profile 319 Aiden KimAutonomous systems meet warehouse reality2017 B GARAGE founded2023 $20M Series ANo GPS · no markers · no prior map

Person / Founder · Engineer · Operator

Aiden Kim Built a Drone for the Places GPS Forgot

The B GARAGE founder spent years studying decisions under uncertainty. Then he turned that training toward a stubborn warehouse problem: knowing what is actually on the shelves.

A warehouse after hours has its own geometry. Tall racks become narrow canyons. Pallets sit above eye level. The database says what should be there, while the shelves hold the less tidy answer. Somewhere between those two versions of reality is the job Aiden Kim chose for an autonomous flying machine.

Kim is the founder and CEO of B GARAGE, a San Jose company that sends drones down warehouse aisles to photograph inventory and compare the physical world with the warehouse management system. The drone is only the most visible part. Behind it sits computer vision, autonomous-flight software, a web application, battery packs and a base station that can swap a depleted battery without waiting for a person.

It is a compact expression of Kim's career. He studied computer science and aerospace engineering at KAIST, earned graduate degrees in aeronautics and astronautics at Stanford, wrote software at Neowiz, Google and Oracle, and researched how machines make decisions when uncertainty is unavoidable. Then he found a market where uncertainty is stacked to the ceiling.

01 / Before the warehouseA career built at the seam

Kim's Korean name is Youngjun Kim. His early professional years moved between software and flight: Neowiz in the early 2000s, a Google internship in 2007, and a senior software engineering role at Oracle beginning in 2010. At Stanford, his academic work made the overlap explicit. His 2016 doctoral dissertation was titled Optimal Planning with Rare Catastrophic Events.

The phrase sounds remote from warehouse counting, but its underlying concern is practical: how should an autonomous system act when the world is uncertain and the costly events may be uncommon? One application in Kim's research involved unmanned aircraft watching wildfires, where useful decisions have to survive incomplete information. His Stanford work also included AI software for a dog-robot research project using hardware developed with Boston Dynamics.

The through-line was not a specific vehicle. It was agency. Kim kept working on the question of what a machine needs in order to operate without a person continuously telling it what to do.

Begins a software engineering chapter at Neowiz.

Works as a software engineering intern at Google.

Joins Oracle as a senior software engineer.

Completes his Stanford doctorate on planning under uncertainty.

Founds B GARAGE around autonomous-flight technology.

B GARAGE did not begin as a warehouse company. Kim first imagined providing autonomous-flight software to drone manufacturers. The team worked around agricultural spraying and infrastructure inspection, both demanding settings. Agricultural work forced attention onto navigation without GPS and avoiding hazards such as power lines. Those capabilities would later travel indoors.

02 / The turnThe customer supplied the coordinates

In 2019, conversations with industry operators sharpened the company. Through a program that matched startups with logistics businesses, Kim met an executive from Kenco Logistics. The executive described a stubborn problem: warehouses are often far from population centers, and inventory checking is repetitive, physical work. Hiring and retaining people to do it was difficult.

The match between capability and need was unusually clean. A warehouse is hostile to the navigation tools that make outdoor drones easy to operate. GPS signals are unavailable or unreliable indoors. High racks create repetitive visual corridors. A useful aircraft must find its way, stop at the right locations, capture readable images and keep going for hours.

“The autonomous drone technology could be highly suitable for B2B logistics warehouse inventory management.”Aiden Kim, 2023

The insight changed the company boundary. Software alone would not deliver the result the warehouse wanted. Full automation required a drone designed for the environment, cameras and onboard processing, a dependable battery system, a ground station, image analysis and a clean connection to the customer's inventory records.

The B GARAGE inventory loop A five-step loop from scheduled flight to autonomous navigation, image capture, warehouse reconciliation and automatic battery swapping. 01 SCHEDULEWeb app setsthe inventory run 02 NAVIGATENo GPS, markersor prior map 03 CAPTURECameras recordthe pallet 04 RECONCILEVision meets WMSinventory records 05 RETURN + SWAPBase station replacesthe depleted battery THE LOOPSTARTS AGAIN
Five moving parts, one outcome: the digital inventory gets another look at physical reality.

This is why B GARAGE became vertically integrated. Kim has said that he expected to sell drone software when he founded the company. The demand for an unmanned service pulled the team into hardware. A flight can be autonomous while its surrounding workflow remains full of manual chores. If someone must place navigation beacons, build maps, transfer files or wait beside a charger, the customer has not escaped the labor loop.

03 / The machineAutonomy is a chain of small permissions

B GARAGE says its camera-vision navigation works without GPS, prior mapping, markers or beacons. The drone processes information onboard in real time, moves across multiple aisles and photographs both high-rack pallets and floor-stacked inventory. The software lets warehouse teams schedule regular cycle counts or request an ad hoc scan, then stores and visualizes what the drone found.

Image quality is part of the autonomy problem. A camera attached to a flying object will vibrate. Kim has described the use of deblurring and other computer-vision methods to make text and labels readable. The point is not simply to return with photographs. The images must be good enough to identify goods and compare them with the warehouse management system.

6-8hReported time for drones to count work that can stretch across weeks manually
99.9%Accuracy Kim reported during the Kenco pilot
$20MSeries A announced in June 2023

Power is the other mundane constraint that becomes decisive at scale. When a battery runs low, the drone returns to its station. The station removes the depleted pack and replaces it with a charged one. This mechanical detail may be less photogenic than autonomous flight, but it separates a demonstration from a scheduled service.

The product is sold through a robot-as-a-service subscription model. That aligns a complicated stack of hardware and software with an operating expense the customer can plan. It also keeps B GARAGE responsible for the system rather than leaving the warehouse to assemble a collection of vendors.

Manual count~95%
Drone pilot99.9%
Accuracy figures Kim described for the warehouse pilot. The narrow-looking gap matters when every miss can distort the operating record.

04 / The companyA garage with two addresses

The name B GARAGE carries two references. It honors Willow Garage, the robotics research company that developed an influential open-source robotics platform. It also remembers B GARAGE's own beginning in a garage. The mix is revealing: institutional ambition paired with a deliberately modest origin.

Kim built across South Korea and Silicon Valley. In a 2024 interview, he described a South Korean team working on the web application's user interface and physical systems including drones and batteries, while the Silicon Valley group concentrated on autonomous flight. That split mirrors his own education: software on one side, aerospace on the other, joined around a single operating problem.

The company announced an 8 billion won investment in 2021. In June 2023 it raised a $20 million Series A led by LB Investment, with participation from Ignite Innovation Fund, Krossroad Partners and existing investor SoftBank Ventures Asia. The round brought reported funding at the time to $30 million and was intended to support engineering, commercialization and team growth.

Funding was a checkpoint, not the operational proof. The more consequential step was turning trials into repeatable work. B GARAGE tested with Kenco Innovation Lab and later described a commercial service in Kenco-operated warehouses. In August 2025, the company said it had supplied its system to a Hyundai Glovis logistics center handling semi-knocked-down automotive parts. That deployment reported 99 percent inventory accuracy and more than a 90 percent reduction in inspection time compared with visual checks.

The shelf is a fact. The database is a claim. Kim's drones keep arranging meetings between the two.

05 / What remainsThe less visible ambition

Kim has discussed extending B GARAGE's software and hardware into ground robots and applying its autonomy stack beyond logistics. The company has also described a larger purpose: making aerial robots ordinary enough to become part of daily life without requiring a pilot. By 2026, Kim was speaking publicly about the wider frame of robots, physical AI and drones.

The practical discipline is what makes that aspiration credible. B GARAGE did not stay with the broadest version of autonomous flight. It moved toward a narrow, recurring and measurable job. It did not stop at navigation. It followed the work through image cleanup, inventory reconciliation, battery replacement and scheduling.

There is an operator's lesson in that progression. A technical capability becomes a business when it inherits the customer's whole annoyance. Warehouse managers do not wake up wanting computer vision or elegant planning algorithms. They want to know what is on the shelf, how certain they can be, and whether the answer will arrive before the next decision depends on it.

Kim's story is therefore less about a drone entering a warehouse than a researcher entering an operation. The rare events of his dissertation gave way to thousands of routine passes down nearly identical aisles. Yet the central question held: how can a machine make useful decisions when it cannot rely on perfect information?

In the warehouse, the answer now flies, looks, returns to charge and starts again.