Grain, sand, feed, biomass - the stuff that arrives in heaps, not on pallets. A Kansas City startup points cameras and drones at the pile and hands you a number you can plan against.
Walk through most of the modern supply chain and you will find a barcode on nearly everything. A pallet of soda. A crate of bolts. A box of phones. Scan, count, done. Then walk out to the yard behind the warehouse, where a hill of grain sits in the open air, and the whole apparatus falls apart. Nobody scans a pile. For decades, the answer to "how much is out there" has been a person squinting at a heap and making a call.
Rebulk, a Y Combinator Summer 2025 company based in Olathe, Kansas, was built to answer that question with a number instead of a guess. The company measures and monitors bulk inventory - feed, sand, grain, logs, biomass, aggregates - using fixed cameras, drones, LiDAR and computer vision. It takes the material that, in the founders' phrase, "doesn't fit in boxes and barcodes" and turns it into a 3D model you can actually plan against.
The reason bulk got left behind is not mysterious. Barcodes and RFID work because a discrete object can carry a label. A pile cannot. You cannot tag a ton of corn stover the way you tag a carton. So the businesses that live on bulk - farms, feed mills, aggregate yards, biomass operators - kept doing it the old way: visual estimates, loader counts, density assumptions, a clipboard and a periodic manual survey.
That works until it doesn't. A miscounted pile means a plant orders material it already has, or runs short in the middle of a run, or carries phantom inventory on its books. Rebulk's own framing of the problem is blunt: most bulk inventory is still managed with "rough estimates, manual checks, delayed updates, and too much guesswork."
The cost of that guesswork compounds quietly. Bulk materials tend to be the raw input at the very front of a production line, which means an error there ripples into everything downstream - procurement, scheduling, cash tied up in stock nobody needed. And unlike a mislabeled carton, a mismeasured pile is invisible. There is no scan that fails, no alert that fires. The number on the spreadsheet simply drifts away from the material in the yard, and usually nobody notices until a truck shows up and there is not enough to load it.
The mechanics are more interesting than "point a camera at it." Rebulk captures overlapping imagery - fixed cameras for indoor piles, drones for outdoor yards, even mobile phone photos in a pinch - and stitches those 2D frames into a 3D volumetric model through photogrammetry. The system identifies the ground surface underneath the pile, which matters because storage yards are rarely flat. A slope you ignore is volume you get wrong.
Then there is the air. A pile of logs or corn stover is not solid - it is material plus a lot of gaps. Rebulk analyzes side-view imagery to estimate the void fraction, the "packing factor," and applies a simple but crucial equation: bulk volume times packing factor equals estimated usable material. That distinction - the difference between how big a pile looks and how much is actually in it - is exactly what a loader-count or a density guess tends to miss.
The choice to mix capture methods is deliberate. A fixed camera mounted over an indoor pile can watch it change hour by hour without anyone lifting a phone. A drone can sweep an outdoor yard too big for any single vantage point. And a plain mobile photo, taken by whoever happens to be standing there, keeps the system usable on a site with no drone pilot and no fixed rig. The output is meant to be the same regardless of how the images arrived: a volumetric model, tied to a location, that can be compared against last week's.
Rebulk's early customers sit at two ends of the bulk world. On one side is agriculture: the company runs a pilot with Cargill and works with one of North America's largest agricultural distributors - the world of feed, grain and the constant question of how much is in the silo yard this week.
On the other side is climate infrastructure. Charm Industrial removes carbon by converting crop residues and wildfire-risk biomass into bio-oil that gets pumped underground. To do that at scale, Charm has to know how much biomass sits in its yards - and biomass comes as irregular log piles and corn stover, about as far from a barcode as inventory gets. Rebulk measured it using drones, fixed monitoring stations and mobile imaging, building 3D reconstructions of piles that let Charm track growth and depletion over time.
Rebulk's system is designed to work in the field, not the lab: remote sites, low connectivity, muddy yards. It produces geotagged, repeatable records that scale across multiple locations - which is the difference between a one-off survey and an inventory system.
Ask who the competition is and the honest answer is not another app - it is the status quo. The clipboard. The loader count. The density assumption. Most bulk operators have never had software purpose-built for their inventory, which means Rebulk's first job is less about beating a rival and more about proving that a measured number beats an estimated one reliably enough to trust.
Directional, not survey data - it reflects the analog methods Rebulk describes replacing, where computer-vision measurement is still the newcomer.
Warren Wang and Cole Robertson met at a Startup Crawl in downtown Kansas City. Wang, the CEO, was a product manager at Milk Moovement, a Series A agtech company, with earlier stops at Dynamo Ventures and Chick-fil-A Corporate. Robertson, the CTO, was head of engineering at the AI startup Chipp and engineer number three at Tropic, which he helped take from seed to Series B, with earlier work at Garmin.
The company they started was not this one. It began as dScribe AI, doing video transcription. The pivot to bulk inventory came later, and the rebrand to Rebulk landed in March 2026 - the new name suggested by Kenneth Bautista, who joined as chief of staff. In a detail that is hard to invent, Wang finished his U.S. naturalization ceremony the same day the company announced the change.
The mix of backgrounds reads like the job description for this particular problem. Bulk inventory is an agriculture-and-industry problem, and Wang came out of agtech. It is also a hard engineering problem - computer vision that has to survive weather, dust and bad lighting - and Robertson came out of shipping AI products and consumer hardware. Neither résumé alone would obviously point at "measure piles of grain." Together they do.
Rebulk sells an inventory operating system: hardware - cameras, drones, LiDAR - paired with the software that turns imagery into volume and weight, delivered to industrial and agricultural operators who manage material across one or many sites. Beyond Y Combinator, whose partner Jared Friedman worked with the team, the company has drawn backing from a cluster of Midwest and specialist investors including KCRise Fund, Abstraction Capital, Flyover Capital, Redbud VC, EquipmentShare and Tekedia Capital. It was named a Kansas City Startup to Watch for 2026 and selected for the LaunchKC cohort.
There is a reason the company is in Olathe rather than the coasts, too. The customers are here - or at least, they are in places like here. Feed mills, grain yards and biomass sites cluster around the middle of the country, not around venture offices, and a team that wants to spend time in real storage yards is better off close to them. Building industrial software from an industrial region is less a constraint than a fit.
The bet underneath all of it is simple and a little contrarian: that one of the least glamorous corners of the supply chain - the pile out back that nobody digitized - is worth building a company around. Rebulk is wagering that when operators can finally see what is in the heap, they will not want to go back to squinting at it.