by Stian Rognlid, Chief Executive Officer, Aquaticode, Norway
Most of the fish that will never reach market are identifiable early on. If given time to assess fish carefully, one at a time, a hatchery operator with thirty years of experience can often tell: a fry that won’t convert feed efficiently, a juvenile whose bone structure is already off, a group that will carry disproportionate mortality into the seawater phase. But at commercial scale, shift after shift, nobody can apply that eye to every fish.
The Decision Farmers Cannot Scale Easily
In Mediterranean seabass and seabream, most of the production cycle is becoming highly instrumented. At the grow-out stage, feed, treatments and harvest are measured to the gram. At the hatchery, the biggest decisions of the whole cycle, which juvenile moves on and which doesn’t, still come down to manual sorting.
A skilled operator can sort by sight, flag the obvious cullable fish and keep the line moving. Decisions beyond that rely on operator experience and they’re hard to keep consistent across a shift. Nobody can phenotype every individual juvenile at one to two grams for growth potential, deformities and survival potential, at commercial speed, shift after shift, with the same consistency at fish number 350,000 as at fish number one.
Experience-based sorting has been the industry’s compromise, a proxy that discards most of the information the fish is already carrying. Buyers of bass and bream require deformed fish to be sorted out as a condition of commercial viability not a welfare gesture.
How Imaging and Deep Learning Transform Decisions
AquaLens brings together three layers: Multi-sensor, multi-angle imaging; a deep-learning stack trained on a large and growing archive of images across species and hatcheries; and soft-touch robotics that move juveniles through the line without the stress of conventional handling.
The imaging captures detail that the human eye and conventional quality control miss at commercial speed. Multiple angles and wavelengths pick up skeletal structure, fin symmetry, pigmentation patterns and subtle early deformities at sizes as small as one to two grams. Deep learning turns that image data into per-fish predictions: This juvenile is likely to grow well in the next stage, likely to carry a deformity that won’t become obvious until the fish is much larger. The robotics then act on those predictions in real time, routing each juvenile to the correct lane.
The result isn’t a faster visual sorter; it’s a different kind of decision. Where traditional assessment produces a size bucket and a cullable pile, AquaLens produces a biological record of every fish that passes through and sorts on that record. Every decision is transparent, consistent and auditable. A head of hatchery can trace why a particular group was routed the way it was and can compare outcomes at sea against the predictions made months before. That feedback loop is how the system gets better with every deployment.
Throughput is the other half of the story. AquaLens sorts up to 350,000 juveniles per day per line. A Mediterranean hatchery can’t afford to slow the line for a research-grade measurement, so the system runs at production speed, non-invasively and slots into existing transfer workflows rather than replacing them.
Early Movers in Bass and Bream
Two early movers are bringing AquaLens into seabass and seabream production: Cooke España and Ilknak.
Cooke España, under the Culmarex brand, is Aquaticode’s development partner for Mediterranean species and has received the first commercial installation. The work with Culmarex is where AquaLens is being built for seabass and seabream. It runs on the core sensing and sorting platform Aquaticode had already developed for SORTpro in salmon, and the deep-learning models train on commercial-hatchery data at one-to-two-gram juvenile sizes. Developing a system with an operator, not to one, is the difference between a technology that works in a lab and a technology that runs through a shift. The partnership was announced in January 2026.
Ilknak, , is the second commercial deployment and the first in Turkey. Aquaticode and Ilknak have agreed to deploy AquaLens across Ilknak’s production. The aim is to replace manual sorting with per-fish biological decisions from the earliest point in the cycle.
Between them, the early work will also generate a biological dataset of the Mediterranean production cycle that no single farm, and no manual sorter, could assemble. For Culmarex and Ilknak the case is operational. For the category, it is the first per-fish biological baseline the industry has lacked the tools to build.
Scaling Hatchery Decisions Where Margins Matter
The Mediterranean bass and bream sector runs on thin margins and high volumes. The levers that move the business are not glamorous. Labour, tank turnover, transfer mortality and the share of a batch that reaches market deformity-free.
Hatchery-stage phenotyping moves several of those at once through a single mechanism: earlier, more accurate culling. A deformed or non-viable juvenile removed at one to two grams frees tank capacity and feed that would have gone to a fish that won’t perform and doesn’t carry forward into transfer mortality or a market-ineligible harvest. The grow-out operator inherits a batch whose expected performance is known beyond what manual quality control alone can predict.
None of this is a new goal. What’s new is being able to run it at the throughput the Mediterranean sector actually requires.
And labour is the lever that is hardest to ignore. Hatcheries across the Mediterranean are struggling to hire and retain the skilled operators that manual sorting depends on, and the shortage is structural, not cyclical. AquaLens shifts the labour profile from subjective sorting at scale to supervising an automated line. That makes the hatchery less dependent on a pool of experienced sorters that keeps shrinking, and less exposed when a shift is short-staffed.
Category Shift Already Emerging in Salmon
The same argument was made in salmon hatcheries a few years ago and the answer is in. SORTpro, Aquaticode’s salmon system, was first installed at AquaChile. Contracted salmon volumes in Chile are above 60 million fish, and the system has since been sold in both Norway and the Faroe Islands.
In salmon hatcheries, SORTpro replaced size-based sorting with per-fish biological sorting. The hatcheries that moved first are no longer asking whether this matters. They’re asking where to deploy the next line.
Bass and bream are now where salmon was when SORTpro was first installed at AquaChile. Different biology, different hatchery layout, different economics. The underlying proposition is the same: Earlier, more consistent biological decisions, at commercial speed, change the math of everything downstream.
