Onur Dogan

I am a Machine Learning Engineer at Lifemote, where I develop AI solutions for telecommunications analytics. I work across the full MLOps lifecycle, from data and model training to deployment, working with LLMs, multimodal models, and time-series models. My work includes productionizing scalable LLM/ML pipelines on AWS and orchestrating large-scale inference.

I completed my MSc in Computer Science and Engineering at Sabancı University under the supervision of Öznur Taştan and Mehmet Keleş from Johns Hopkins University. My thesis, CoFINE: Hierarchy-Aware Semi-Supervised Learning for Fine-Grained Behavioral Classification, focused on learning from limited labelled data in fine-grained and hierarchical classification settings.

Before Lifemote, I worked on computer vision for robotics, with a focus on robotic perception. My work included real-time segmentation, automated data annotation, simulation-based data collection with NVIDIA Isaac Sim, and building perception pipelines for real-world robotic systems.

I also have a research background in computational biology, where I worked on graph and ML methods for RNA-seq, single-cell RNA-seq, and protein sequence data.

My experience spans computer vision, multimodal learning, time-series modelling, LLMs, and production ML systems. I am especially interested in building reliable learning systems and developing methods that make AI systems more robust, adaptive, and trustworthy in real-world settings.