The data scientist who declared war on the checkout line - and is winning, one billion transactions at a time.
At some point around 2001, Abhinai Srivastava was doing work at Yahoo that didn't have a name yet. "Data scientist" wouldn't enter the mainstream vocabulary for another decade. He was just a guy from IIT Delhi who was very good at finding patterns in things - and had joined a mid-sized internet company before anyone knew what the internet would become.
He would spend the next decade refining that skill at an uncommon scale. At Microsoft, he wasn't just writing algorithms - he was building the foundations of how the world's largest software company would charge for attention. The relevance engine he helped create at adLab determined which ads showed up when someone searched on Bing. Then he built the first version of the Bing Knowledge Graph, the semantic layer that lets a search engine actually understand what you're asking about. He was, without fanfare, doing the kind of work that gets quietly embedded into billions of daily interactions.
Then Facebook, where he applied the same instinct to social signals. A recommendation algorithm he built reduced spam pages on the platform by 60% - a number that, at Facebook's scale, translates to hundreds of millions of cleaner user interactions. He was, by this point, one of the most quietly accomplished ML engineers in Silicon Valley. The resume would have funded a comfortable career indefinitely.
Instead, he called his old IIT Delhi dormmate, Mukul Dhankar - who had separately been thinking about a very specific problem: the 30 to 40 minutes a day he spent waiting in his corporate cafeteria line. Dhankar had been working on computer vision at Toyota. Between the two of them, they had the full stack: the AI expertise and the vision technology. What they needed was a problem worth building for.
They chose the most boring, universal, and friction-saturated moment in retail: paying for things. In October 2012, Mashgin was born.
The bet was audacious in a particular way. Self-checkout had existed for decades and was widely considered a technology that worked - not great, but functional. The standard approach used barcodes: an item goes under a scanner, a laser reads a stripe, a database matches it. Abhinai and Mukul decided to skip all of that. No barcodes. No scanners. No specific orientation required. The machine would look at whatever you put on the platform and figure it out, the way a sharp-eyed cashier would - but in half a second.
Getting the machine to see that well took years. The proprietary hardware uses multiple 3D cameras to build real-time spatial models of each item. The AI has been trained on over 100,000 common items. And when it encounters something new, it can learn it in 30 seconds. The system now operates at 99.99% accuracy - a number that, in checkout terms, means it almost never makes a mistake.
The business case arrived gradually, then all at once. In 2022, Mashgin raised $62.5 million in a Series B led by New Enterprise Associates, valuing the company at $1.5 billion. Fast Company ranked them #3 Most Innovative Company in Retail. Forbes put them on the AI50 list. And Alimentation Couche-Tard, the Canadian conglomerate that owns Circle K, announced a deal to roll out Mashgin's technology across more than 7,000 stores globally.
The numbers from deployments are the kind that make procurement teams cancel other meetings. The Denver Broncos saw concession sales rise 34% per game. UT Austin's stadium posted a 125% revenue increase. DK Stores cut lines by 67%. In each case, the mechanism was the same: when a line disappears, people buy more. Abhinai has a line for this. "We understand that 75% of retail is still offline. When retailers use our technology, in many cases the sales go up by a huge margin just because there are no lines anymore."
By March 2025, Mashgin was processing 40 million transactions per month - up from 3 million just three years earlier. That is 1,233% growth. In 2024, the system handled 400 million transactions. The cumulative total crossed one billion. The machine Abhinai and Mukul built in Palo Alto is now embedded in convenience stores, sports stadiums, airports, university dining halls, ski resort lodges, hospital cafeterias, and corporate campuses across multiple countries.
Abhinai manages all of this from 849 East Charleston Road in Palo Alto, with a team that is conspicuously small for the scale of what it operates. When a new colleague asked how the engineering team had already deployed 5,000 kiosks and tens of millions in ARR, the answer was essentially: we built it right the first time, even when building it right was slow. "Focus on doing things the right way, even if it's slow" is the version of that philosophy that Abhinai has committed to as a motto.
The sectors still ahead of him are the ones where barcode-based checkout never made it to begin with. Hot food. Multi-item grab-and-go. Hospital cafeterias where you're juggling a tray and a phone. Ski resort lodges where people want to spend thirty seconds on lunch and four hours on the mountain. These are the environments where Mashgin's technology doesn't just improve checkout - it makes it possible at all.
"Computer scientists have long dreamt of the possibilities of what artificial intelligence could help humankind achieve. Today, Mashgin helps lead the way to realize that dream in the retail sector by giving people back precious time."
- Abhinai Srivastava, Founder & CEO, Mashgin
MONTHLY TRANSACTIONS (APPROXIMATE) - 1,233% GROWTH OVER 3 YEARS
Most startup founders talk about finding product-market fit fast. Abhinai spent two decades doing something different: making himself genuinely hard to replicate. The career that preceded Mashgin wasn't a stepping stone. It was an apprenticeship in how AI actually works at scale - not in theory, not in demos, but in systems serving hundreds of millions of people daily.
At Yahoo, he was exploring data before the tooling existed to explore it cleanly. At Microsoft, he was building relevance algorithms for advertising when paid search was still figuring out its own rules. The Bing Knowledge Graph - the semantic layer that powers structured search understanding - had his fingerprints on its first version. These weren't side projects. They were load-bearing pieces of infrastructure for one of the world's largest technology companies.
Facebook came next, where the scale got bigger and the problem shifted from commercial relevance to social integrity. A recommendation algorithm he built cut spam page prevalence on the platform by 60%. At Facebook's scale, "60% less spam" is an intervention that affects hundreds of millions of people's daily experience - the kind of result that appears in no one's press release but echoes through the product for years.
By the time Mashgin needed someone who understood how to build AI that works in the messy physical world - not the clean data world - Abhinai had done exactly that work for two decades. The dormmate phone call to Mukul was less a pivot than a deployment.
"We understand that 75% of retail is still offline. When retailers use our technology, in many cases the sales go up by a huge margin - just because there are no lines anymore."
- Abhinai Srivastava
#3 Most Innovative Company in Retail - joining Stripe, Canva, Microsoft, and SpaceX on the 2022 list.
Named to Forbes' annual selection of the most promising private companies in North America utilizing artificial intelligence.
Won the Gold Edison Award for the AI-Powered Touchless Self-Checkout System - tech's equivalent of the Oscars for innovation.
Mashgin was accepted into YC's Winter 2015 batch - one of the most selective programs in startup history.
$1.5B valuation achieved in May 2022 via $62.5M Series B led by NEA. Total funding: $74.7M.
Crossed the cumulative 1-billion-transaction milestone, with 400 million of those occurring in 2024 alone.
The Mashgin kiosk doesn't look like a checkout machine. It's a countertop device with multiple 3D cameras pointing at a platform. You put your items down. It looks at them. In half a second, it knows what you have. The full checkout - items recognized, payment accepted, receipt issued - takes about 5 seconds.
The architecture underneath is a specific triumph: the system builds real-time three-dimensional models of each item from any angle, without requiring consistent packaging, specific orientation, or visible barcodes. It has been trained on 100,000+ items. When it encounters something new, a staff member can teach it in 30 seconds.
What makes this commercially significant isn't just the speed - it's the locations it unlocks. Stadium concessions, hospital cafeterias, ski resort lounges, corporate dining halls - these are environments where barcode-scanning self-checkout was never going to work. A burger doesn't have a SKU. A cafeteria tray of mixed items can't be individually scanned. Mashgin handles these contexts that traditional retail technology treats as unsolvable.
The result is 99.99% accuracy - exceeding human cashier performance - at a speed that turns "I'll grab something quick" from an intention into an actual behavior.
"Focus on doing things the right way, even if it's slow."
"We understand that 75% of retail is still offline. When retailers use our technology, in many cases the sales go up by a huge margin just because there are no lines anymore."
"Computer scientists have long dreamt of the possibilities of what artificial intelligence could help humankind achieve. Today, Mashgin helps lead the way to realize that dream in the retail sector by giving people back precious time."
"I am super excited about the potential applications of Computer Vision in real world."
Abhinai didn't jump into hardware AI at 25. He spent two decades building the exact skills Mashgin needed - relevance, recommendation, spam detection - before applying them to a physical problem.
He was doing data science before the profession had a name, at Yahoo in 2001. The field caught up to where he already was. He simply kept building.
Mashgin built proprietary hardware (not an iPad hack) to support advanced AI capabilities. That combination - custom device plus advanced ML - is the moat that keeps competitors at bay.
Abhinai had the data science. Mukul Dhankar had the computer vision. They met at IIT Delhi, separated for a decade, and reassembled in Silicon Valley with complementary superpowers.
His operating principle - do it right even if it's slow - shows up in 99.99% accuracy and kiosks that essentially never need maintenance calls. The numbers downstream justify the patience upstream.
Self-checkout was considered "solved." Abhinai identified that barcode-based checkout had never worked in stadiums, hospital cafeterias, or ski resorts. He went there first.
Mashgin can learn a brand-new item it's never seen before in just 30 seconds.
The Denver Broncos saw concession sales jump 34% per game after installing Mashgin.
UT Austin's stadium posted a 125% revenue increase with Mashgin deployed.
Abhinai recently took up running and has achieved a sub-6-minute mile.
He started working in data science in 2001 - nearly a decade before "data scientist" became a recognized job title.
Mashgin has saved the equivalent of 30+ years of cumulative human time that would have been spent waiting in checkout lines.
The spam-reduction algorithm Abhinai built at Facebook cut bad pages by 60% - at a scale of billions of users.
Mashgin integrates with 50+ payment networks, POS systems, and loyalty platforms out of the box.