ABOUT ARPANET
The sovereign AI company: our models, our gateway, your control
Arpanet Bilişim A.Ş. is a vertically integrated AI company based in Ankara. We build the layers most vendors rent — five in-house model families and one OpenAI-compatible gateway — and put seven products on top, so organisations can run advanced AI inside their own perimeter. Founded in 2018 inside ODTÜ Teknokent, we ship a platform engineered for the KVKK from the first line of code — and ready for the EU market it serves, where the AI Act now reaches Turkish vendors too.
- 2018
- Founded
- ODTÜ Teknokent
- Ankara
- 5
- Model families
- 7
- Products
- 43+
- Model providers
SINCE 2018
From a research group to an integrated AI platform
We did not start as a thin wrapper over someone else's API. We started with the hard parts — speech, language, vision — and kept building until the whole stack, from the models to the gateway to the products, was ours. That is the arc, and it is why the promises we make on sovereignty and control are ones we can actually keep.
Arpanet is founded as a machine-learning research company inside ODTÜ Teknokent in Ankara — one of Türkiye’s leading technology parks — focused on speech, language and vision.
The first products solved discrete problems: high-accuracy speech-to-text, translation, sentiment analysis, face and image recognition, and large-scale media monitoring.
Those capabilities matured into five in-house model families — Verinova for language, Blab for voice, Solab for speech-to-text, Spook for vision and OCR, and VeriRAG for retrieval.
Qevron unified them behind one OpenAI-compatible endpoint — first-party models plus 43+ external providers, with routing, caching and cost analytics.
Arpanet ships seven products across two legal entities — Arpanet Bilişim A.Ş. in Türkiye and Arpanet BV in the Netherlands — so customers can choose the jurisdiction their AI runs in.
RESEARCH & ENGINEERING
Built by people who build the hard parts
The intelligence under our products is developed in-house, not licensed. That is what lets us make promises other vendors cannot — on where your data lives, on what the models can do, and on how the whole system behaves.
Deep R&D roots
Our research began in 2018 inside ODTÜ Teknokent in Ankara — one of Türkiye’s leading technology parks — supported by TÜBİTAK-backed R&D programmes, and R&D remains the core of how we work.
In-house model families
Five families cover language, voice, speech-to-text, vision and retrieval — trained and served by us, tuned for production rather than rented per call.
Our own product brands
Qevron, Calleague, Blab, Talky, SetGet, Sosyabot and Pickerman are our own registered product trademarks — not third-party tools rebadged.
Protected, evolving IP
Our core methods are the product of years of patented and patent-pending R&D — the platform is ours to protect, and ours to keep advancing.
DEPLOYMENT & SOVEREIGNTY
Your AI, inside your perimeter
Because we own the models and the gateway, where they run is your decision, not ours. The same platform deploys three ways — and for organisations under the KVKK, and for anyone serving EU users where the AI Act now applies across the border, that choice is increasingly a legal question, not only an architectural one.
The full stack runs inside your data centre. Data never leaves the building; you hold the keys, the logs and the audit trail.
A dedicated, network-segregated environment with controlled egress — the platform runs in your own tenancy, not a shared multi-tenant service.
A managed deployment in the region you choose when you want speed over sovereignty — still your data boundary, still behind your SSO.
One honest distinction: 'isolated' means a segregated network with controlled egress; 'air-gapped' means no route off the box at all. They are different risk profiles, and we will always tell you which one you are buying.
HOW IT FITS TOGETHER
Our models. Our gateway. Your AI.
Everything sits on two owned layers. That vertical integration is the point — nothing in a product is a black box bolted on from outside.
Five in-house families — language, voice, speech-to-text, vision and retrieval. The intelligence, owned.
Qevron: one OpenAI-compatible endpoint to our models plus 43+ providers, with routing, caching and cost analytics.
Seven products on top, from voice agents to media intelligence — each drawing on the same models and gateway.
WHAT WE DO
From requirement to a design that keeps scaling
We work with organisations to define what they actually need, identify the right solution, and design systems that stay open to growth and change. Through strategic partnerships with leading manufacturers and service providers, at home and abroad, we make sure the efficiency, security and performance of what we deliver returns on the investment behind it.
Understand the requirement
We start from the problem, not the product — scoping what a real deployment needs before anything is built.
Choose the right solution
The right model, the right deployment, the right integration — matched to your constraints, not our catalogue.
Design for growth
Architectures that stay open to expansion and new capability, so today's build is not tomorrow's rewrite.
Deliver measurable value
Efficiency, security and performance that make the investment pay — backed by strategic partnerships across the sector.
Build on the platform beneath the products.
Bring your workload to a sovereign stack you fully control.