Sorting used textiles with AI, NIR, and a recycling framework
A circular economy project with Wargön Innovation, funded by Vinnova: image recognition and near infrared spectroscopy fused into an automated sorting process.
Machine learning that ships. Sovereign by design.
We build and deploy production ML for Swedish industry and the public sector, on infrastructure you own, or on sovereign cloud. Your data never leaves the perimeter.
On-prem & edge capable
EU jurisdiction & data residency
No vendor lock-in
Trusted by
Wargön InnovationCircular economy
ItmotiveConsulting
Braha ITConsulting
Every engagement ends in a system your team runs
Open-weight models, nothing behind someone else's API
Read the work →
By design
Numbers we hold ourselves to
100%
of systems designed to run on infrastructure you own
0
bytes of your data required to leave your perimeter
EU 1
jurisdiction, one legal framework, fully auditable
100%
source code and documentation handed to your team
How we engage
Start small.
Start small.
Own everything.
A staged ladder, not a leap of faith. Each stage is fixed in scope, ends in something you keep, and earns the next, starting with a two-week sprint at a published price.
01Sprint
ML Feasibility Sprint
2 weeks · från 95 000 krA fixed-scope audit of your data, infrastructure, and use case. We tell you honestly whether ML is feasible, what it would return, and what it would take, before you commit to anything bigger.
02Pilot
Pilot
4–8 weeks · fixed quote after the SprintOne use case, taken from the roadmap to a working prototype on your data, evaluated against a success metric we agree on in writing before we start.
03Build
Production build
Scoped per projectWe deploy, integrate, and harden the system on infrastructure you control: monitoring, rollback, and reproducibility from day one, with compliance evidence produced as we ship.
04Operate
ML Ops & improvement
Monthly retainerDrift monitoring, retraining, and continuous improvement, while we train your team to run it. Success is your engineers not needing us on the phone.
Proof of work
How we build,
How we build,
documented.
We're a young studio. As client projects close, they appear here as named case studies with verified numbers. Until then, our work notes show exactly how we think and build.
Building Qonverge, our AI talent matching platform
The talent matching platform we built for ourselves and now run as a product, in production at Itmotive and Braha IT. What we built, and where we stopped the AI.
A reference architecture for sovereign ML
The stack we reach for when a system has to run on-prem for a decade: open-weight models, portable serving, and an exit path at every layer.
Where we work
Two verticals,
Two verticals,
known deeply.
We'd rather understand two industries properly than ten superficially, and both of ours run on trust, jurisdiction, and uptime.
Manufacturing & industry
Visual inspection, predictive maintenance, energy optimisation: machine learning that runs on the factory floor, OT-safe, with no round-trip to anyone's cloud.
Edge-first: we quantise and deploy models on-device (Jetson, STM32, RISC-V), built for offline-first operation.
Public sector
Machine learning for organisations that answer to citizens: data residency in Sweden, GDPR and EU AI Act compliance designed in from the first commit, and systems your own staff operate.
Built for public scrutiny: every model documented, every decision auditable, nothing hidden behind a vendor's terms.
Why sovereignty
Convenience is rented. Control is built.
The easy path routes your most valuable asset, your data and the models trained on it, through infrastructure and APIs you don't own. Quince builds the other path.
The default
Rented intelligence
- Data and prompts leave your perimeter to third-party clouds and model APIs.
- Pricing, rate limits, and model behaviour change underneath you.
- Compliance depends on a vendor's terms you can't fully audit.
- Switching provider means rebuilding. The lock-in is the product.
The Quince way
Sovereign intelligence
- Models run where your data lives: on-prem, private cloud, or at the edge.
- An open-weight stack you can inspect, fine-tune, and keep running indefinitely.
- Compliance designed in: GDPR and the EU AI Act as architecture, not afterthought.
- Portable by default: no single provider you can't walk away from.
The studio
Nordic engineering,
Nordic engineering,
applied to your data.
Quince AI is a machine-learning studio in Gothenburg, on Sweden's west coast, a region that quietly built world-class automotive, manufacturing, and deep-tech engineering. We bring the same discipline to AI: rigorous, dependable, and built to last.
Start the conversation
Let's build AI you don't have to hand over.
Thirty minutes, no commitment. Tell us the problem and the constraints, and we'll tell you honestly whether, and how, machine learning fits.