It is 12:18 p.m. at a Circle K outside Phoenix. A construction worker walks in with two energy drinks, a breakfast burrito, and a bag of jerky. He sets them on a small countertop with four cameras pointing down. A screen lights up. The total appears. He taps his card. He walks out. The whole transaction has taken eleven seconds. There is no cashier. There is no line. There is, somehow, no barcode scanning either. There is just Mashgin, doing the quiet thing it has been doing in 800-plus locations across America while the rest of retail tech argued about itself on LinkedIn.
I // Who they are nowThe unicorn that nobody talks about
Mashgin is a computer-vision company. That is a polite way of saying it has spent a decade training machines to look at a tray full of randomly arranged food and merchandise - hot dogs, croissants, ski-mitten-warmer packets, three different sizes of Red Bull - and ring them all up at once. No scanning. No "please move the item." No frowning cashier.
It works in stadiums. It works in airports. It works in cafeterias that feed 4,000 lawyers a day. As of its last public count, the company had processed more than 35 million transactions and pushed over $310 million in sales across its kiosks. In May 2022, NEA led a $62.5 million Series B that valued the company at $1.5 billion. And then, in true Mashgin fashion, the company went back to shipping kiosks and stopped issuing press releases.
II // The problem they sawThe line was always the product
For most of retail's recent history, the checkout line has been treated as a fixed cost of doing business. Customers tolerated it. Operators staffed it. Cashiers, bless them, endured it. And then a peculiar thing happened: human attention got shorter, labor got more expensive, and the act of standing behind another human while they hunted for a barcode on a bag of pretzels started to feel anachronistic - somewhere between rotary phones and writing checks at the grocery store.
The traditional self-checkout was, charitably, a half-answer. It moved labor to the customer without removing the friction. You still had to find a barcode. You still had to argue with the machine about an unrecognized item in the bagging area. The line, once again, was the product. Mashgin's founders looked at this and concluded that the problem wasn't the cashier. The problem was the very idea that a single item needed to be picked up, oriented, scanned, and placed back down.
III // The founders' betFrom IIT Delhi to a Toyota cafeteria
Mukul Dhankhar and Abhinai Srivastava met as undergrads at IIT Delhi, sharing a dorm before scattering into separate careers. They reunited in Silicon Valley. Mukul was working on computer vision for Toyota - teaching cars how to interpret the world through cameras - when the bet crystallized in the least glamorous setting possible: the corporate cafeteria. Lunch was taking him 30 to 40 minutes a day, almost entirely spent in line. If a Toyota could be taught to recognize a stop sign at 60 miles an hour, the thinking went, surely a computer could be taught to recognize a sandwich sitting still on a tray.
The bet was technical and contrarian. At the time, computer vision was busy chasing autonomous vehicles and content moderation. Pointing it at a corporate cafeteria sounded mundane. It also sounded like a market nobody else was looking at. They went through Y Combinator, picked up early backing from Matrix Partners, and started building a kiosk - in hardware, in software, in the deeply unsexy work of getting one machine to do one job extremely well.
IV // The productFour cameras and a habit of seeing
The Mashgin kiosk is, physically, an unremarkable rectangle - a screen, a payment reader, a tray-sized surface ringed by 3D cameras. The interesting part lives in the model. Place anything down: a banana, two pastries, a tallboy, a sandwich still in its triangular plastic prison. The kiosk identifies each item simultaneously, totals them, and waits for payment. Throughput is roughly 10x faster than barcode checkout. Operators report fewer abandoned carts, shorter peak-hour lines, and reduced shrink.
The system also learns. Operators can add new menu items - a seasonal pastry, a regional brand, a custom snack at a ski lodge - by showing the kiosk a small number of examples. The model picks it up in minutes. This is the kind of detail that sounds modest until you remember that "training a deep-learning system on a new SKU" is the sort of phrase that, until recently, required a research lab.
What you can actually do with it
If you're a stadium operator, you stop losing concession sales to halftime lines. If you're a c-store, you free your one available employee to actually help customers instead of standing behind a register. If you're a hospital cafeteria, you let nurses on a 12-hour shift get food in 20 seconds. If you're a ski lodge, you let a skier in heavy gloves grab snacks and leave without fumbling for a wallet. Mashgin doesn't replace hospitality; it removes the chore.
V // The Mashgin TimelineA decade of removing one minute at a time
Mukul Dhankhar and Abhinai Srivastava co-found Mashgin in Palo Alto.
Y Combinator. First Matrix Partners check. Kiosk v0 takes shape.
Series A. Early traction in corporate cafeterias and stadium concessions.
Pandemic puts "touchless checkout" in every operator's RFP overnight.
$62.5M Series B led by NEA. $1.5B valuation. Unicorn status.
Multi-tray functionality launches at Palisades Tahoe ski resort.
Quietly rolling kiosks into airports, c-stores, and arenas at scale.
VI // The proofThe customers no one expected
Press releases are easy. Deployments are not. Mashgin's customer list reads like a who's-who of operators with no patience for vendor theater. Aramark, Compass Group, and Sodexo - the three food-service giants that quietly run the cafeterias inside America's offices and hospitals - all use Mashgin kiosks. HMSHost runs them inside eight U.S. airports. Madison Square Garden uses them on concourses where one slow transaction can mean missing a tip-off. Circle K has rolled them into convenience stores. BP's ampm did the same. Palisades Tahoe set them up so skiers could check out without removing their gloves.
Below the celebrity names, the numbers tell the story.
Mashgin, by the numbers
VII // The missionMake the line stop existing
Most companies are vague about their mission. Mashgin is unusually specific: remove the line from physical retail. Not optimize it. Not gamify it. Not migrate it to a mobile app. Remove it. The premise is that humans are bad at queuing, machines are now good at seeing, and the only reason we still wait for a cashier to scan a coffee is that nobody, until recently, had built the kiosk that made it unnecessary.
There is a quieter idea underneath - that retail labor, when freed from the act of ringing items, can do the thing humans actually do better than machines, which is be hospitable. A cashier who isn't a scanner can be a host. A barista who isn't a register operator can be a barista. This is the unglamorous, slightly humanistic argument that Mashgin makes by accident every time one of its kiosks frees up a minute of someone's shift.
VIII // Why it matters tomorrowThe compounding minute
There are roughly a billion physical transactions a day in U.S. retail. If you remove a minute from each of them, you have given back the equivalent of roughly a million human work-years a year. That is not a marketing claim. That is just arithmetic. Mashgin is not going to capture all of those transactions. It does not need to. It needs to keep capturing the ones where the line is the most expensive thing in the building - the stadium at halftime, the airport between flights, the hospital cafeteria at 7 a.m. - and keep training the model on the next ten thousand SKUs.
The competitive landscape is real. Amazon's Just Walk Out is the headline rival. Trigo, AiFi, and a few others are circling the same idea from different angles. Several have shut down. Mashgin's bet has been narrower and quieter: not the whole store, just the checkout. Not surveillance, just a kiosk you walk up to. It has turned out to be the bet that scaled.
IX // Closing sceneBack at the Circle K, eleven seconds later
The construction worker is back in his truck before the door has finished closing behind him. The next customer is already at the kiosk. Two energy drinks, a coffee, a packaged donut. The cameras see what he sees. The total appears. He taps. He leaves. The clerk, freed from the register, is restocking the cold case. The line that used to define this hour of this store does not exist, because the kiosk in the corner is quietly running an algorithm that a decade ago was being trained, in part, on a Toyota cafeteria sandwich.
Mashgin built a unicorn out of one annoying observation: lines are a tax, and the tax was optional. The market is catching up to that idea slowly. The kiosks are catching up to it eleven seconds at a time.