Founder’s note
Let your cameras stop recording and start deciding.
The observation we started from in 2021 was simple: almost nobody watches what security cameras record. Footage is opened after something has gone wrong, to look back. The camera sees everything, but nobody is told what it saw.
Twenty years in enterprise IT taught me that new technology only creates value once it is connected to the systems already in place. That is why we built CX Teknoloji around the cameras a site already has, not around selling new ones. The only hardware we install is the server that processes the image.
Today we are a team of more than ten at GTÜ Technopark in Gebze: computer-vision and deep-learning engineers, software developers and integration specialists who do the installations. I run product, engineering and sales. The people who build the models and the people who install them sit at the same table, so every false alarm from a pilot reaches the person who wrote the model and the rule.
Abdulaziz Akyol · Founder & CEO
- 23ready-made solutions
- 5solution categories
- 8sector packages
- 10+people on the team
- <40 msdetection per frame
- 30day pilot
How it works
From pixel to decision
The stream from an existing IP camera passes through five stages on a GPU server at the site. Every stage is measurable; the result is stored as a live alarm, a dashboard metric and an evidence clip.
- Camera streamRTSP · H.264 / H.265, existing camera
- Decoding25 fps · hardware-accelerated on GPU
- Detection and tracking<40 ms per frame · deep learning
- Rule enginezone + time window · a single frame never raises an alarm
- Event and dataalarm · dashboard · MES, with an evidence clip
What we do
All solutionsOne vision stack, five problem areas
Every solution runs on the same inference pipeline. A new requirement therefore means a new rule and model, not new hardware. Where no ready-made solution fits, we build project-specific video analytics.
Site and perimeter security
Production and operations
Retail analytics · EscherSense
Industrial quality inspection
Sector packages
From the field
What the camera tells us
How we think
Five principles, the same on every installation
Existing cameras first.
Selling cameras is not our business. Our solutions work on the IP cameras already on site; the only hardware we add is the processing server.
A single frame never raises an alarm.
A situation becomes an event only after it is confirmed over a defined time window. False alarms wear out the operator’s trust, so we design against them from the start.
We record the event, not the person.
No face recognition, no personal records; processing is anonymous. The system produces the time, place and evidence clip of an event, not a profile of a person.
Measurement, not promises.
We agree before the pilot what it will measure. For thirty days false alarms are reviewed daily; the decision is made on results measured on your site.
Data stays on site.
By default video is processed on a server inside the facility and raw video does not leave. Cloud processing is possible if you choose it.
How we work
From discovery to roll-out
- camera and goal
Discovery
We assess your existing cameras free of charge and agree which event will be measured where.
- one week
Installation
Camera streams are connected to the server; zones, thresholds and time windows are drawn on your site.
- 30 days
Pilot
False alarms are reviewed daily; the written criteria are measured on site and reported.
- integration and new sites
Roll-out
Events are connected to your MES, ERP and VMS; the proven set-up moves to other lines and plants.
Results
CustomersMeasured on the customer’s own site
vehicle and equipment safety: vehicle–pedestrian proximity detection on existing cameras
City Lojistik
counting accuracy: people counting, dwell time and conversion
Emilio
output per shift: line balancing from station-level measurement
Textile manufacturer · anonymous
Customers · City Lojistik · Emilio · Suwen · Saat & Saat · Turkcell · Gallery Crystal
Two brands, one technology
Writing
CX Teknoloji blogWhat I write at CX Teknoloji
Investment, cost, regulation and strategy articles I signed on the CX Teknoloji blog are also published here (in Turkish).
- Kamerayla OEE mi, PLC/MES ile OEE mi? Karar Rehberi
- Görüntü İşleme Firması Nasıl Seçilir? Kriterler ve 14 Soru
- Yapay Zekâ Kamera Fiyatları Neye Bağlı? Maliyet Kalemleri
- Ajan Tabanlı Yapay Zekâ: Uyarı Vermekten Karar Almaya
- Işıksız Fabrika Mümkün mü? Otonom Üretimde Görüntünün Rolü
- Kalite 4.0: İstatistiksel Kontrolden Sürekli Muayeneye
- Doğru Pilot Nasıl Tasarlanır? 30 Günde Karar Verilebilir Sonuç
- AB Yapay Zekâ Yasası (AI Act): Fabrikada Kamera Analitiği
- Veri Yönetişimi: Analitik Verisi Kimin, Nerede, Ne Kadar Süre?
- Kurulum Süreci: Keşiften Devreye Almaya 6 Adım
- Analitik Olgunluk Modeli: Kurumunuz Hangi Basamakta?
- İnsan ve Yapay Zekâ İş Birliği: Operatörün Yerini Almak Değil, Gözünü Büyütmek
- İSG'de Dijital Dönüşüm: Reaktiften Proaktif Güvenliğe
- Görüntü Analitiği Yatırım Getirisi: ROI Nasıl Hesaplanır?
Field notes
FAQ
Frequently asked questions
What is CX Teknoloji?
CX Teknoloji is a computer-vision and AI company at GTÜ Technopark in Gebze, Türkiye. It turns the IP cameras that factories, warehouses, ports and stores already have into systems that produce decisions for occupational safety, site security, production efficiency and quality inspection. It offers 23 ready-made solutions in 5 categories and 8 sector packages.
Who founded CX Teknoloji and when?
Abdulaziz Akyol founded CX Teknoloji in 2021. He is the company’s founder and CEO and previously was Head of IT at Civil Mağazacılık (2012–2021).
Do I need new cameras?
Usually not. The solutions work on the existing IP camera streams (RTSP, H.264/H.265). The only hardware installed is a GPU server that processes the video. In-line industrial quality inspection uses a dedicated industrial camera on the line.
Does CX Teknoloji use face recognition?
No. Processing is anonymous and no personal records are kept. The system records the time, place and evidence clip of an event, not a person’s identity. By default video is processed on a server inside the facility.
How does a project start?
With a free camera assessment. Installation takes about a week, followed by a 30-day pilot in which false alarms are reviewed daily and agreed criteria are measured. The roll-out decision is made on those measured results.
Let’s look at your site together
A 15-minute call is enough to work out which of your existing cameras can measure which event.