Co-Founder & CTO
Caner Onat
AI engineer. Leads AI development, R&D, software engineering and technical operations at Piksel Bayt.
Works on machine learning, computer vision, large language models and software architecture. Responsible on Piksel Bayt's R&D projects for the technical process, from problem definition and feasibility through model development to the system running in the field.
Writes the research notes on this site, which describe the technical questions the team meets in projects and how it approaches them.
Areas of work
- Machine learning
- Computer vision
- Large language models
- Software architecture
- Model operations (MLOps)
Posts
- Why Turkish evaluation sets are a problem in their own rightWhy we build our Turkish evaluation sets from scratch, the fundamental problems with machine-translated sets, and how domain terminology distorts measurement.
- Before putting agents into production: safety guardrailsA practical framework for where to place human approval in tool-using agents, how to define permission boundaries, and reversibility.
- Vision models at the edge: the latency–accuracy trade-offWhich accuracy loss is acceptable when moving a detection model to real time on constrained hardware through distillation, quantisation and TensorRT.
A technical assessment for your AI project
Your project's feasibility, risks and timeline are assessed in a technical consultation.