Mashgin

Mashgin makes a self-checkout kiosk that uses computer vision and machine learning to identify items placed on a tray, removing the need to locate and scan barcodes. Multiple cameras build a three-dimensional understanding of each item, allowing several products — including non-packaged goods such as produce or prepared food — to be recognized at once and rung up in a single transaction. The company states identification accuracy of 99.9% (99.99% on its solution overview page), item recognition in under half a second, and complete transactions in roughly 5 to 10 seconds, which it describes as about four times faster than a cashier or traditional point of sale.

The kiosks are deployed in convenience stores, corporate cafeterias, hospital and healthcare cafes, university campuses, sports stadiums and arenas, airports, ski resorts, grab-and-go micromarkets and cafe dining. The platform includes cloud reporting, item tagging and grouping, an API for data access, and integrations with third-party systems; installations are described as requiring only power and an ethernet connection, with a recommended countertop footprint of 50" x 30" and setup possible within an hour. Named venues and customers include Madison Square Garden, AT&T Stadium, Petco Park, SAP Center, Lumen Field, Angel Stadium, Circle K, Stinker Stores, United Dairy Farmers, Charlotte Douglas and St. Louis Lambert airports, Stanford Medical Center, Penn State, Arizona State University and Texas A&M, alongside foodservice operators Aramark, Compass Group and Sodexo.

Founding story

Mukul Dhankhar, then working on camera-based perception for Toyota, spent 30-40 minutes a day waiting in his corporate cafeteria line and began testing whether computer vision could speed checkout. He flew to California and reconnected with Abhinai Srivastava, a classmate from the same dorm at IIT Delhi; an early demo convinced Srivastava to join. The pair cold-called executives at large corporate cafeteria operators, got enthusiastic responses, and built a prototype. They applied to Y Combinator days before the Winter 2015 batch began while raising a seed round, and the batch turned the project into a company.

Business model

Kiosks are sold as a service rather than as a hardware purchase. The recurring service fee bundles the kiosk and peripherals, Mashgin's proprietary computer vision software, cloud hosting and administration, ongoing software updates, menu assessment, management training, system monitoring and support, and reporting with basic data API integration.

Subscription-style, per-kiosk service fees covering hardware, software, cloud hosting and support; pricing is quoted on request. The company said in May 2022 that it was already profitable.

Traction

The company reports more than 5,000 deployed locations, over 1.5 billion transactions processed and more than 40 million monthly users, including over 100 sports stadiums. At the May 2022 Series B it reported over 800 locations, more than 35 million transactions and $310 million in transaction volume processed to date. Customer-reported results include throughput increases of 3x to 4x and sales gains ranging from roughly 18% in corporate cafeterias to 400% at some venues; the Denver Broncos saw 96% faster throughput and 34% more transactions per game at Mashgin locations.

Latest developments

The company's site reports more than 5,000 locations, 1.5 billion transactions used to train its models and over 40 million monthly users, with published customer results from venues including VyStar Veterans Memorial Arena (Legends Global), Palisades Tahoe, Honda Center, UT Austin and St. Joseph's Hospital. It is recruiting senior software engineers for product and computer vision/deep learning roles in Palo Alto under a hybrid remote model.