The company teaching ordinary cameras to read the world - turning images into operational data.
Tiliter's product-recognition engine at a supermarket self-checkout, Sydney. The camera identifies loose produce with no barcode, no sticker and no lookup screen - a result in under a second. Photograph illustrates the company's core technology.
Barcodes were meant to fix the checkout. They never covered the messy edges: loose apples, bulk nuts, unpackaged bread - the things a scanner cannot read. In 2017, three engineers in Sydney - Marcel Herz, Martin Karafilis and Christopher Sampson - decided the camera should do the reading instead. After years of computer-vision research, they founded Tiliter to recognise products with no barcode at all.
The first product was deceptively simple: point a small camera at whatever a shopper places on the scale, and identify it - a banana, a bunch of silverbeet, a bag of oranges - in under a second, with no sticker and no lookup screen. That single trick, done reliably and at retail scale, became the wedge that opened one of Australia's largest grocers and, from there, an international footprint.
Today Tiliter has widened its aim well beyond the checkout. Its platform of AI Vision Agents takes any image and returns structured data: validating an expiry date, counting boxes on a pallet, spotting a crack in a surgical tool, reading a label, or scoring how clean a floor is. The company's one-line summary - "turn images into operational data" - is also the whole thesis.
Tiliter builds AI-powered verification workflows: capture a photo, apply an automated visual decision, trigger an action. Its customers are retailers and enterprises that live and die by visual work nobody enjoys doing by hand. In grocery, that means recognising unbarcoded products at self-checkout and cutting the friction of manual code entry. Named retail customers include Woolworths Group, with the technology running across more than 1,000 stores in Australia and New Zealand, alongside Countdown in New Zealand, Netto Marken-Discount in Germany and Westside Market in New York.
Beyond retail, the same engine reaches into logistics, manufacturing, healthcare, construction and other industrial operations - counting parts, inspecting for damage, reading documents and verifying that tools and inventory are where they should be. The problem it solves is consistent across all of them: visual checks are slow, subjective and hard to scale when a human eye is the only sensor.
Turn images into operational data.
Tiliter ships its capabilities as modular agents - small, composable AI tools you point at an image - plus workflow apps and a streaming layer that runs on existing cameras.
Identifies products, parts, packaging and assets - including unbarcoded produce - in under a second with high accuracy.
Bolts a camera and software onto existing scales and checkouts to auto-identify and weigh items, removing manual entry.
Label Validator, Text Extractor, Receipt Processor, Object Counter, Damage Detector and Cleanliness Evaluator - all in one library.
Connects live IP and CCTV cameras to Vision Agents for real-time analysis - no new hardware, no rip-and-replace.
Turns site photos into an objective 1-to-5 cleanliness score across multiple locations, replacing subjective inspections.
Extracts text from documents and packaging, validates expiry dates and serial numbers, and digitises receipts.
Cashierless and computer-vision retail is a crowded field - Amazon's Just Walk Out, Trigo, Grabango, Standard AI, Mashgin and Shekel all chase versions of the frictionless store, while broader inspection platforms like Landing AI target industry. Tiliter's difference is where it starts: rather than rebuilding a whole store, it retrofits onto scales, checkouts and cameras retailers already own. That lowers the barrier to adoption - and it means every deployment feeds a data engine that processes tens of millions of images a year.
Figures reflect Tiliter's publicly stated metrics and product claims. Bar lengths are illustrative of relative positioning, not precise benchmarks.
Tiliter operates as a B2B SaaS and technology-licensing business. Early revenue came from camera-and-software systems retrofitted onto point-of-sale hardware; the newer motion is software-only - Vision Agents and Vision Stream that run on a customer's existing cameras, plus workflow apps like Cleensight sold on a subscription basis. The company reports estimated annual revenue in the region of US$10M and a compact, research-driven team.
The expertise is deep computer vision: recognising objects that resist traditional automation because they have no barcode, no fixed shape and no label. Co-founder Martin Karafilis has been recognised on the Forbes 30 Under 30 list, and the founding team spent years in R&D before commercialising - a heritage that shows in the platform's focus on real-world deployment over demo-stage novelty.
Marcel Herz, Martin Karafilis and Christopher Sampson launch the company after years of computer-vision research.
The camera-and-software system begins identifying and weighing unbarcoded produce at self-checkout.
Investec Emerging Companies leads the round, with Eleanor Venture and Cornell University participating.
Product recognition expands across Woolworths and international grocers including Countdown and Netto.
Tiliter reframes around modular Vision Agents, plus Vision Stream and the Cleensight cleanliness app.