At NTNU, I worked as a machine learning lead on an interdisciplinary sleep-science project, translating clinical questions into ML problems and presenting results to a non-ML audience. I built a cross-modal pipeline to predict circadian phase across 35,725 study participants, replacing a $500-per-subject lab-based procedure.
For long sensor signals, I built a masked autoencoder that cut effective sequence length by up to 10x while matching or beating state of the art, and reduced training time by up to 7x. I also reduced model-update compute by 90% with a drift-triggered retraining rule in an online Case-Based Reasoning system.
On NTNU’s SmaRTWork project, I was tech lead and full-stack developer for an ML-based app for back pain, developed in collaboration with the Norwegian Labour and Welfare Administration. I refactored and deployed the system as a patient-facing service behind a 300-patient clinical trial, and remained responsible for it in production.
At Anansy, I built and shipped a symbol-based communication app for autistic children, combining LLM speech transcription with embedding-based symbol retrieval. I own its full stack and deployment, and the app was piloted with 10 speech therapists. I’m the author of five peer-reviewed papers, including one spotlight, and my PhD thesis has been submitted.
