Artificial intelligence is the umbrella term for methods that let machines perform tasks that normally require human intelligence — seeing, understanding language, deciding and learning. In organisations today it is used mostly as machine learning, deep learning and large language models (LLMs).
Work in this area
Since 2021 Abdulaziz Akyol has been running AI in the field at CX Teknoloji — on real cameras and production lines. His articles cover the practical side of moving models from the lab to the field, measuring accuracy and aligning AI with enterprise IT governance.
- Video-analytics models running in the field with up to 96% accuracy (CX Teknoloji)
- Field notes on model drift, retraining and dataset discipline
- Enterprise RAG, AI agents and MCP architecture
- AI projects aligned with the EU AI Act and data-protection law
Frequently asked questions (3)
Where should a company start with AI?
With a measurable business problem and a single use case whose data is ready. A project chosen with a value × feasibility matrix and taken to production within 90 days builds learning and trust before any large platform investment.
Why do AI projects fail in the field?
Usually because of data and environment, not the model: lab data does not represent field lighting, angles and variability; accuracy is not measured on site; and models drift away from the data over time. Field calibration and continuous measurement close that gap.
What does Abdulaziz Akyol do in AI?
At CX Teknoloji, which he founded, he builds computer-vision and IoT solutions that turn existing IP cameras into real-time decision systems, with 23 ready-made solutions in five categories across retail, safety, manufacturing and logistics.
- Accelerating edge AI models: ONNX, TensorRT and INT8 quantizationArtificial intelligenceExport a PyTorch model to ONNX and speed it up with ONNX Runtime and TensorRT: FP16, INT8 calibration, accuracy loss and p50/p95 latency, with working code.
- AI agents and MCP: architecture, security and a first project for companiesArtificial intelligenceAI agents and MCP explained: connecting them safely to ERP and CRM with least privilege, prompt injection defences, approvals, audit logs and a first project.
- An AI roadmap for CIOs: your first production project in 90 daysLeadershipA first AI production project in 90 days for CIOs: use-case selection matrix, data readiness, build-or-buy, 30-60-90 day plan, metrics, governance and budget.
- Enterprise RAG: how to build an AI assistant that talks to company documentsArtificial intelligenceBuild a RAG assistant that answers from company documents with citations: parsing, chunking, embeddings, pgvector, hybrid search, permissions, evaluation.
- EU AI Act: a practical compliance guide for Turkish companiesArtificial intelligenceWhen does the EU AI Act apply to Turkish companies? Risk classes, GPAI, the timeline after the Digital Omnibus, video analytics and a compliance checklist.
- Video analytics on existing cameras: the chain from RTSP stream to eventComputer visionReal-time analytics on the IP cameras you already have: RTSP intake, GPU decoding, detection and tracking, a rule engine and event output, in five links.
- Getting people counting to 96%: field calibration notesComputer visionA model that scores 99% in the lab drops to 85% in the field. Camera angle, counting-line position, night data and measurement method close the gap to 96.2%.
- Combining IoT and camera analytics: MQTT and an event-driven architectureIoTA camera is the richest sensor. On one event bus with PLC signals, door contacts and temperature it yields new data: event schema, MQTT topics, lessons.
- Keeping detection models alive in the field: drift, retraining and dataset disciplineArtificial intelligenceTraining is the short part. The hard part is keeping a model accurate in 40 locations as season, light and layout change: drift, dataset and versioning.
Older technical notes (HTML5, PowerShell, Windows administration, Active Directory, HP/Aruba switches) are available in Turkish only. Türkçe →