A supermarket can sell you three kinds of apple and then ask you to become an apple expert before it will let you leave. Somewhere inside a self-checkout menu, your fruit awaits its formal introduction. Christopher Sampson and his co-founders thought a camera ought to be able to handle that part.
It was a small irritation with a large technical question attached. Could a machine recognise loose produce, without a barcode doing the introductions? Tiliter, the company Sampson co-founded in 2017, grew out of that question. Its early task was wonderfully concrete: help a shopper buy what was already sitting on the scale.
Today Sampson is Tiliter’s co-founder and co-CEO, based in Sydney. His route there runs through electrical engineering, established companies, a garage and overseas customers. The interesting turn comes after the invention. Getting software to see an object was one challenge. Getting another person to believe in the business required him to leave some comfortable engineering habits behind.
The circuit before the checkout
Before the supermarket came the guitar. Sampson’s interest in music developed into an interest in electrical engineering. He grew up in country New South Wales, has Wiradjuri and Kamilaroi ties, and completed his engineering degree at the University of Newcastle. Education was familiar territory early: as a small child, he attended a lecture while his mother was training as a teacher.
His formal training continued with a Master of Engineering Science in Systems and Control at UNSW, following his Newcastle honours degree in electrical engineering. His early work included undergraduate engineering at Rio Tinto, an acoustic engineering role and a progression at Cochlear from summer intern to graduate engineer and test systems engineer. He had worked inside organisations with existing systems before trying to create an organisation of his own.
The sequence gives his later advice to graduates a practical basis. Sampson encourages them to take time in established businesses and different roles. “Take the opportunities when they come but don’t jump into something before you’re ready.” His own career offered several opportunities to learn before the company began.
A high-school friend, a university connection
Tiliter’s founding team brought together relationships formed at different ages. Sampson and Martin Karafilis had known each other since high school. Sampson met Marcel Herz while studying at UNSW. The three founded Tiliter in September 2017 and entered the UNSW Founders accelerator network.
The university connection mattered beyond the introductions. Tiliter became part of the 10x Accelerator’s 2018 cohort, joining a setting where technical founders could work on the commercial questions around an invention. Mentors, investors and corporate strategy entered the picture alongside the recognition system.
There is something pleasingly ordinary about the object at the centre of all this effort. An apple has no obligation to resemble the photograph in a catalogue. For a shopper, recognising it can be almost automatic. For a checkout system built around labelled packages, loose produce presents a different job. Sampson’s team chose a problem people could understand without first sitting through a lecture on artificial intelligence.
When the prototype needed a passport
The first commercial opening came from overseas. Herz’s German connections helped the team reach a major point-of-sale manufacturer. It paid for an early prototype, made using 3D printing, and brought the founders over to work with its engineers. For a young business, a buyer willing to work with a prototype was a consequential kind of encouragement.
Publicity provided another opening. A national news report showed Tiliter’s technology operating in a store, and enquiries followed from retailers and technology companies. Sampson has described the discomfort of putting himself and the product in public: “As an engineer it was a little bit uncomfortable for me.” The experience became part of his advice to other founders: approach journalists and show people what you have built.
A demonstration has an advantage over an abstract promise. A retailer can watch a familiar transaction and judge whether the proposed change makes sense. Tiliter’s subject was already part of the buyer’s working day. The founders still had to win attention, but once they had it, they could point at a scale.
“Compute support helps you build, but customers prove the value.”Christopher Sampson, August 2026
The apple joins an existing system
By February 2020, Tiliter was testing its technology with Woolworths. The retailer’s Scan & Go app already let shoppers scan packaged products and pay using their phones. Loose produce was the troublesome part. Tiliter supplied recognition technology around scales and cameras to help bridge that gap.
That is a useful detail in Sampson’s story. The business could enter through a specific transaction inside an existing shopping system. A shopper did not need to understand the model, and a retailer could consider a recognisable operational problem: how to handle goods that arrived without a convenient code.
In October 2020, Tiliter announced a $7.5 million funding round led by Investec Emerging Companies, with participation from Eleanor Venture and Cornell University. At the time, early customers included more than 20 Woolworths stores in Australia, Countdown in New Zealand and US retailers including New York’s Westside Market. The company described plans to expand in Europe and the United States.
The geographical spread makes the early German opening easier to appreciate. International work had become part of the company’s actual customer base. Sampson was helping build a business whose software needed to operate far beyond the room where its engineers could watch it.
The unglamorous art of keeping it running
In May 2021, writing about Tiliter’s deployment infrastructure, Sampson said the company had customers on four continents. His account concerned software containers, remote updates and the ability to roll a device back to an earlier version. These are modest subjects beside an invention’s debut, but they reveal what happens when a prototype acquires customers.
A device in another region needs a manageable way to receive changes. Sampson explained how Azure IoT infrastructure helped Tiliter distribute software and handle updates without repeatedly moving every component. The rollback mechanism mattered too: an earlier version could be restored without downloading all of its data again.
There is a different kind of ambition here from the stage demonstration. A demonstration earns attention for a few minutes. Deployment engineering serves the people using the product afterward. Sampson’s technical writing puts him inside that ongoing work, explaining the machinery that lets a growing company support its installations.
He also discussed the retail application in a 2021 video conversation with Ruth Yakubu and technical product manager Paulwyn Devasundaram. Nuts, fruit and bakery products featured among the examples. The everyday shopping list remained a useful way into a subject otherwise crowded with specialised language.
A partnership beyond the model
In November 2024, Tiliter announced a partnership with SAPP International. The division of work was explicit: SAPP would support production scaling and hardware capability, while Tiliter would supply its AI software. Hardware, maintenance and support were part of the proposition Sampson described.
The announcement reported Tiliter’s AI scale solution in more than 1,100 Australian Woolworths locations, alongside its presence in New Zealand. That was a company-reported deployment figure at that date. It concerned systems supporting tasks including price checking, e-commerce and Scan & Go, rather than a count of entirely automated supermarkets.
For Sampson, the partnership placed another practical consideration beside recognition accuracy. Retailers have equipment to install and maintain. An AI company entering that world has to account for the physical life of the product as well as what its software can identify.
Why the interface keeps returning
Sampson had already put user experience at the centre of his thinking in an essay published in October 2022. He compared AI’s development with computing before personal computers became commonplace. Computing power could exist before people had a sufficiently useful way to interact with it. He argued that AI faced a similar delivery problem.
It is an argument that suits the checkout. The recognition model works behind an interface, and the shopper meets the interface. A capable system can still create a laborious transaction if its results are difficult to use. Sampson’s essay described an engineering team spending effort on how it delivered technology as well as how it improved recognition.
By January 2025, he was announcing Tiliter’s Vision AI Agents in connection with the NRF retail event. During 2026, his updates described a further step: applications built around the agents for customers who lacked software or workflows into which they could integrate them. CleenSight, an application for assessing cleanliness, was one example he named.
The company now presents agents for tasks such as counting objects, extracting text and checking labels. Those are broader assignments than identifying produce. Yet the delivery question remains familiar: how does a result from an image become something a person can use in a working day?
Another kind of recognition
In August 2026, Sampson returned publicly to the difficulty of winning Australian customers. Early overseas business, he explained, helped establish credibility at home. He argued for companies willing to try locally developed technology and emphasised the value of an early customer’s real operating experience.
His position carries the memory of selling something before the company behind it was familiar. A customer contributes revenue, but also a place where a product has to work. That experience gives the next buyer something more concrete to consider. It is a form of recognition that no camera can supply.
Sampson’s career has moved from learning inside established engineering organisations to helping build one that other businesses depend on. The scale, the remote update and the application interface each demand attention. Somewhere near the beginning sits a shopper searching a screen for fruit, giving an engineer a very reasonable question to ask.