Julia Romero, PhD

I recently completed my Ph.D. in Computer Science at the University of Colorado Boulder. I am interested in AI research regarding computer vision, vision-language (VLMs) and multimodal sensor models, foundation models, interpretability, and model adaptation and transfer. My doctoral research focuses on effective and efficient ML/AI for sensor data that handles data and system constraints, spanning applications in ubiquitous computing, video understanding, and Earth Observation. My research has spanned these topics:

• Vision & Geospatial Foundation Models: I develop and interpret self-supervised vision foundation models for Earth Observation tasks, examining pretraining and transfer learning/adaptation strategies. For FM model development, I implemented and maintained multi-cluster distributed computing infrastructure on an NSF supercomputer (before LLMs knew Slurm...) [in submission, 2026]
• Multimodal Video Understanding: I build graph-based vision models for fine-grained activity recognition that are trained on diverse data but infer from a single egocentric (smart-glasses) input [ICCVW'25]. This work with Intel Labs won 1st place in the Ego-Exo4D challenge (out of 20+ teams) and led to an invited spotlight talk at the CVPR 2025 Egocentric Vision Workshop.
• Human-Centered Sensing Systems: I design sensing algorithms for resource-constrained wearables and ML pipelines for health data, spanning physiological monitoring with smart earbuds [HotMobile'24], injury-risk modeling with fitness wearables through internships, activity recognition with motion sensors [Sensors'23], and large-scale analysis of physical activity behavior [IJERPH'22].

My Ph.D. research was funded in part by an NSF IUCRC award from a proposal I co-wrote, which supported a collaboration with Intel Labs from 2023 to 2025. Previously, I was a research intern at Nokia Bell Labs in Cambridge, UK, and worked on wearable computing and health ML at Stryd, Kinesis Integrated, and the Johns Hopkins Applied Physics Laboratory.