Computer vision and neural networks
In-house detection and recognition models: plates of 100+ countries, faces, people, vehicles, speed, dimensions. Trained on our own data, inference on GPU and edge.
PAG is an Azerbaijani IT and AI company. We build our own technology from the neural network to the board at the network edge: computer vision engines, plate and face recognition, Velum Vision video analytics, automation platforms, mobile apps and hardware integrations. Our systems run 24/7 on 40+ sites in Azerbaijan and in pilots in Georgia and the Czech Republic.
In-house detection and recognition models: plates of 100+ countries, faces, people, vehicles, speed, dimensions. Trained on our own data, inference on GPU and edge.
Smart CCTV: Face ID, ANPR radar, tracking, footfall counting, heat maps, real-time events on any IP camera.
Backends, web panels, operator workstations, billing, reports, API. High-load services in Node.js, Python, Java, C#, TypeScript.
iOS and Android: PAG Mobile for drivers, Appina for service teams, resident apps for housing complexes. Published on the App Store and Google Play.
Mini-PCs and single-board computers at the barrier, controllers, Dahua/Hikvision, payment kiosks, POS terminals, MikroTik networks and VPN.
Payment systems (Epoint, bank cards, Apple/Google Pay), access control, 1C and ERP, Telegram and WhatsApp bots, voice assistants, AI operators.
Describe the process you want to automate or the data you need to recognise.
In 1–2 weeks we build a working prototype on your data and show the metrics.
Go live, 24/7 support, staff training, development along your roadmap.
…and more sites across the country. The list grows every month.
Computer vision (PyTorch, ONNX, TensorRT, OpenCV), Node.js, Python, Java, C#, TypeScript, Kotlin and Swift for mobile, PostgreSQL and SQLite, Linux and Windows edge devices, Cloudflare and MikroTik for networking.
Yes. We take on projects that need AI, computer vision, video analytics, automation and hardware integration. Write to [email protected].
Yes. We trained our own plate recognition models for Azerbaijan, Georgia, the Czech Republic and other countries, and we train models for client tasks the same way.
In Baku. Computer vision, software and hardware engineers work locally — same-day site visits.
Own products live on 40+ sites, an in-house recognition network at 99.6%, a full cycle from hardware to mobile app, and 24/7 support.
Write to us — we will model the economics of your site and come back with a financial model within 1 day.