ML engineer & pipeline builder based in Copenhagen.
I build machine learning systems, computer vision, image restoration, and data pipelines. Background in media production gives me hands-on context most engineers don't have.
I'm a Copenhagen-based freelancer working at the intersection of machine learning and media production. My focus is computer vision, image processing pipelines, and applied ML, built and shipped in Python.
I have a background as a film and TV production technician, I understand the workflows and problems I'm building tools for. Currently into adversarial ML and model optimization.
Real-ESRGAN fine-tuning pipeline for manga scan restoration. Custom loss function, 9000+ training pairs, 60k+ iterations. Open source.
Quantization research on Llama-3.2-1B using SpinQuant. Exploring how rotation-based quantization affects model behavior and output distributions. Ongoing.
Exploring goal-preservation and simulated functional emotional states in language models — whether affect-like signals improve alignment and long-term context coherence. Early stage.