Visual intelligence at scale
A high-performance engine that produces and reads high-end images across millions of files.

AI that sees our world- and knows what it means -
Today's AI is powerful, but it recognizes without understanding — and it is confidently wrong exactly when the stakes are highest. We build visual AI that grasps context and meaning, so it stays dependable for the decisions that actually matter.
AI recognizes without understanding — and it is confidently wrong exactly when the stakes are highest.
We build visual AI that grasps meaning, so it stays dependable for the decisions that actually matter.
[ what we build ]
A high-performance engine that produces and reads high-end images across millions of files.
More capability per compute: we push past the limits on image size while cutting time, cost and footprint.
AI that reads context and meaning and can explain why, so it stays accurate where ordinary models hallucinate.
The industry's answer to every problem is the same: make it bigger.
But size does not fix the flaw that matters most to anyone betting on AI: today's models are confidently wrong.
Our bet is different. We give AI a real grasp of context and meaning, so it stays reliable on the hard cases, adapts without months of retraining, and never trades accuracy for a fluent answer. That is the line between an impressive demo and a system you can actually deploy.
It holds up on the rare, high-stakes cases that quietly break ordinary AI.
Update its behaviour without retraining a giant model, at a fraction of the cost.
It sharpens meaning without distorting what is actually there. Trust by design.
[ efficiency & responsibility ]
Most AI improves the brutal way: by burning ever more compute, with the cost and the carbon to match.
One large AI model can cost a year of supercomputer time.
We made the opposite choice in our compute and in our algorithms. More capability per watt, past the limits on image size and in a fraction of the time. Efficiency is not only margin; it is how we keep AI light on the planet.
[ our edge ]
Our advantage is not a single feature. It is more than a decade of R&D in distributed computing and AI, and a deep grasp of how machines represent meaning. We publish openly to set the standard and earn credibility, while the hardest parts stay ours.
It is the kind of edge that is hard to copy and grows over time: every problem we solve makes the next one easier.
For the technically inclined: read the research[ industries ]
Understanding pays off wherever a wrong call is expensive and meaning is not obvious from the image alone. We start with luxury, a demanding, high-margin proving ground, and the same engine extends across markets worth far more.
Reading the codes and context of a brand enables recommendation, tagging and discovery that feel expert, not automated.
Scenes where the unusual is not, by default, an anomaly to be corrected.
Recognition where a mistaken context carries real cost.
Reading configurations that are rare yet entirely valid.
Telling a legitimate deviation from a true defect.
[ deployment ]
Control and confidentiality are not afterthoughts. Run our technology inside your own environment, or consume it as a service. You decide where the intelligence lives.
[ company ]
soixante-dix [ cs ] is an independent European DeepTech, founded and led by women, on a clear mission: build visual AI the world can trust. We pair frontier engineering with deep research to close the gap between AI that looks impressive and AI that is dependable in the field.
No black box. Every decision can be traced, inspected and explained.
Efficient by design in our compute and our algorithms to stay light on the planet.
Founded and led by women, rare in frontier AI by design. Different builders ask different questions.
Lauréat BPI Émergence
DeepTech
10+ years of R&D
Founded & led by women
[ get in touch ]
Partners, customers, investors: if you are betting on AI that is actually dependable, we would like to hear from you.
hi@70-cs.ai