Edgify’s $9M Turns Store Devices into Shared Edge Brains

Edgify raised $9 million in Series A+ funding from Rank Ventures and Mangrove Capital Partners, lifting its total capital to $25 million as it scales an edge AI platform already live on more than two thousand grocery stores.

The London company turns existing cameras, scales, self-checkouts and point-of-sale terminals into a coordinated network that learns locally and shares insights without shipping raw data to the cloud. That model is now leaving grocery for convenience, restaurants, apparel and industrial sites.

The bet is straightforward. Dense fleets of devices already sit on retail floors. Edgify treats that installed base as the training surface rather than as a set of dumb endpoints that must phone a distant server.

Devices That Learn Together Inside the Store

Edgify’s software sits on hardware retailers already own. It is hardware-agnostic and works across makers including Zebra Technologies and Bizerba. Each device trains models on the data it sees, then federates the learned patterns across the store network so a new scan-avoidance trick caught at one till can appear at every other till without the footage ever leaving the building.

The company says this cuts cloud infrastructure spend and latency while keeping customer and operational data inside the four walls. On its site the platform currently shows these live figures:

• 2,042 stores
• 9,437 devices
• 400M+ samples
• 98.58% precision

Those four numbers sketch the operating picture. Thousands of stores supply the volume. Nearly ten thousand devices supply the local compute. Hundreds of millions of samples supply the training signal. Precision above 98 percent supplies the commercial case for leaving the system running.

Core use cases start with loss prevention:

• Product recognition without barcodes
• Scan avoidance and product switching detection
• Missed items in carts
• Back-of-house waste tracking

A testimonial on the company site from Richard F. Webber, director of information technology at Indiana Grocery Group, notes immediate benefits on self-service scales and “really impressive” accuracy.

The approach runs models directly on edge devices rather than relying on constant cloud round-trips or new dedicated servers. Shared learning still happens. Raw video and transaction streams do not have to travel to do it.

Nine Million Dollars and Familiar Backers

Rank Ventures and Mangrove Capital Partners backed the Series A+ round. Mangrove is a returning early backer Mangrove from the 2020 seed. Rank had already led a seed extension in autumn 2022 after following the company for years.

On its portfolio page Rank calls Edgify the builder of the world’s most effective serverless MLOps platform description using a federated machine-learning framework that keeps data on low-powered edge devices. Managing partner Rajan Dosanjh said the winners in AI will own the point where data is created and that Edgify’s traction with major grocers and Zebra shows the market is ready.

Co-founder and CEO Nadav Israel framed the raise around a simple conviction: intelligence should live where data is created and devices should learn as one. COO Mitchell Goldman added that retail packs smart devices into a small space with strict cost, latency and privacy demands, making it the ultimate testing ground.

The fresh capital will accelerate grocery rollouts with existing partners and fund the push into new verticals. The raise sits among other European AI infrastructure raises that also bet on physical-world data.

Familiar backers matter here. Continuity from seed through seed extension into Series A+ signals that the same investors still see the federated edge thesis holding as the store count climbs past two thousand.

From Failed Photo App to Store-Floor Network

Edgify did not start in retail. Co-founder Nadav Israel previously built Pixoneye, which analysed personal photo galleries on smartphones. When that business struggled, the Israeli engineering team repurposed the edge-computing core for training deep-learning models directly on hardware customers already owned.

1. 2018-2019, Pivot from Pixoneye photo analytics; Edgify launches focused on edge training.
2. 2020, $6.5 million seed from Mangrove Capital Partners, Octopus Ventures and an unnamed semiconductor strategic investor.
3. Autumn 2022, Rank Ventures leads seed extension after multi-year diligence.
4. August 2026, $9 million Series A+ takes total funding to $25 million (€21.6 million).

The funding path compresses into a short ledger of rounds and totals:

The company is headquartered in London with deep Israeli technical roots. Today the public face includes CEO/CTO Nadav Tal Israel and COO Mitchell Goldman.

The pivot itself is the mechanism story. On-device photo analysis and on-device retail vision share the same constraint: limited local compute, sensitive personal or commercial data, and little appetite for constant cloud uploads. Grocery simply offered denser hardware and clearer loss-prevention math than a consumer photo app ever did.

How Edgify Cuts Against Cloud-Heavy Rivals

Computer-vision loss prevention is crowded. Trigo and AiFi build systems aimed at cashierless stores that usually need dedicated infrastructure. Everseen specialises in checkout analytics. Edgify’s claimed difference is avoiding new servers and long installs by using spare compute on devices the retailer has already paid for.

Whether the lighter footprint becomes a durable moat or simply a cheaper on-ramp is still being tested at larger scale. Retailers already facing high shrink and labour costs have a clear incentive to try anything that plugs into what they own.

The competitive contrast is less about model accuracy alone and more about what must be installed before the first inference runs. A system that rides Zebra and Bizerba hardware already on the floor shortens the path from pilot to multi-store rollout. Rivals that need new cameras and backend stacks face a longer capital conversation with the same buyer.

Retail Was Only the Proving Ground

Grocery delivered dense device fleets, clear ROI in loss prevention, and strict privacy rules that reward keeping data local. That combination let Edgify prove federated edge learning in the wild. The same friction appears wherever fleets of cameras, sensors and terminals meet the physical world.

Near-term expansion targets include:

• Convenience stores
• Quick-service restaurants
• Distribution centres
• Apparel retail

Longer term the company points to transportation, logistics, manufacturing and warehouse operations. In each case the pitch is the same: turn isolated legacy hardware into a real-time network that operates independently of the cloud.

On X, early posts already frame the story this way, noting the jump from 2,000-plus grocery stores into restaurants, apparel, warehouses and manufacturing rather than treating it as a pure shrink product.

Grocery remains the reference deployment. Convenience and QSR add speed and smaller footprints. Distribution and apparel add different SKU mixes and back-of-house flows. The orchestration layer is what the company wants to carry across those settings without rewriting the core idea.

What the Numbers Say About the Larger Bet

Edgify targets a retail computer vision market it puts at $15.8 billion while citing broader edge AI growth from roughly $36 billion to approximately $386 billion by 2034. Independent edge AI market growth forecasts show different absolute figures (one recent Fortune Business Insights range runs higher), yet the direction is consistent: processing is moving closer to the point of data creation.

That shift carries second-order effects. Retailers keep sensitive footage and transaction streams inside the store, reducing compliance surface and bandwidth bills. Hardware OEMs such as Zebra and Bizerba gain a software layer that makes their installed base smarter without forklift upgrades. Cloud-centric computer-vision vendors face a cheaper, privacy-friendlier alternative that can still share intelligence across a fleet.

The same pattern is visible in other European AI infrastructure deals that treat the physical world as the next training and inference surface. Parallel rounds in defence simulation and AI security show capital still flowing into specialised infrastructure even inside a crowded field.

One such deal is the other European AI infrastructure raises that also bet on synthetic or physical environments. Another sits in the crowded AI security funding race where specialised layers keep attracting capital.

Investors Back Learning at the Source

The Series A+ thesis tracks a single line from Dosanjh and Israel alike. Value accrues to whoever owns the moment data is created. Federated learning on low-powered store devices is the mechanism they are funding to capture that moment without a cloud detour.

Rank’s multi-year watch before leading the 2022 seed extension, then returning for the A+, reads as conviction in that mechanism rather than a quick momentum trade. Mangrove’s path from 2020 seed to this round adds the same signal from the earliest institutional check.

Retail’s cost, latency and privacy constraints made the proving ground harsh on purpose. Goldman’s point is that a platform which survives those constraints has a template worth carrying into other dense device environments. The $9 million is meant to fund that carry, not only thicker grocery coverage.

European AI infrastructure capital has been willing to fund specialised layers that sit close to messy physical or operational data. Edgify’s raise fits that pattern: less general-purpose model training, more orchestration of hardware the customer already bought.

How Shared Patterns Travel Without Raw Footage

The store-floor loop is simple to state and strict in practice. A till or scale sees a new behaviour. The local model updates on that device. Learned patterns, not the underlying video, move across the store network so every other till can recognise the same behaviour.

That split is what lets the company claim both fleet-wide intelligence and data that never leaves the building. Cloud spend falls because training and inference lean on spare local compute. Latency falls because decisions do not wait on a round-trip. Compliance surface shrinks because sensitive streams stay inside the four walls.

Hardware agnosticism keeps the loop portable. Zebra and Bizerba gear already on site can join the same network without a single-vendor rip-and-replace. For a retailer measuring shrink against labour and install cost, that portability is part of the pitch alongside the 98.58 percent precision figure.

Scale is the open test. The public baseline of 2,042 stores and 9,437 devices shows the loop running in grocery. Convenience, restaurants, apparel and industrial sites will show whether the same pattern-sharing holds when layouts, device mixes and staff workflows change.

The Template Leaves the Aisle

Edgify now has the capital to push the coordinated-device model beyond grocery. The company reports it is already live with retailers across the United States and Europe. Its public metrics of 2,042 stores and 9,437 devices give a concrete baseline against which the next wave of convenience, QSR and industrial deployments can be measured.

Israel’s founding line remains the spine: intelligence should live where data is created, and devices should learn as one. Retail proved the mechanics. The $9 million is the fuel to test whether the same mechanics hold when the devices sit on factory floors and warehouse racks instead of checkout lanes.

If they do, the real product was never just better loss prevention. It was a reusable edge orchestration layer that turns any dense device fleet into a shared brain that never needs to phone home with the raw footage.

Grocery wrote the first chapter with clear ROI and harsh constraints. The next chapter asks whether that chapter was a niche win or a portable template. The fresh capital, the returning backers and the live store count are the instruments the company will use to answer it.