Denisa Qori McDonald
Machine Learning Researcher and Engineer building robust learning systems for real-world sequential data
I design machine learning systems for structured, human-generated signals, with a focus on recovering reliable structure under noise, sparsity, limited observability, and distribution shift.
My work has been grounded in specific domains: knitted textile sensors, conversational behavior, and physiological time series. Those domains are particular, but the problems they surface are not: sparse observations, cross-user generalization, real-world variation, and representations that remain stable outside controlled conditions.
The underlying questions in my work — representation quality, generalization under shift, and learning useful structure from difficult signals — also connect directly to broader modern machine learning, including foundation-model-oriented work.
What I Work On
Robust representation learning
Designing features and learned representations for sequential and structured data that remain informative across users, environments, and real-world variation. The goal is to recover stable structure from noisy, sparse, or partially observed signals.
Temporal modeling under real-world variation
Developing machine learning approaches for behavioral, physiological, and sensor-based time series, combining signal processing, feature construction, and learned models to capture meaningful temporal structure across users, settings, and conditions.
End-to-End ML Systems
Building complete pipelines from raw sensor inputs and preprocessing through modeling, evaluation, and deployment. I treat architecture, evaluation design, and implementation constraints as core parts of the machine learning problem.
Research Perspective
I am motivated by a central question: how do learning systems remain reliable when data are noisy, users differ, and real-world conditions shift?
Across projects, I treat predictive performance as only part of the goal. I also ask whether the learned structure is stable, interpretable, and likely to transfer beyond the original setting. That perspective shapes how I think about representation quality, robustness, and evaluation under real-world variation.
Selected Work
Representation Learning for Sparse Textile Sensing
Developed machine learning methods for touch localization and gesture recognition from minimally wired knitted sensors, showing how robust representations and temporal models can recover human intent under hardware, physical, and interaction constraints.
This work reflects broader machine learning challenges in limited observability, representation learning, and cross-user generalization from sparse measurements.
Computational Modeling of Naturalistic Social Behavior
Built interpretable temporal representations of head movement and social behavior in face-to-face interaction, enabling quantitative modeling of clinically relevant behavioral structure in ecologically valid settings.
More broadly, this work addresses how subtle, noisy temporal patterns can be modeled in ways that remain both informative and interpretable.
Physiological Time-Series Modeling
Developed early machine learning pipelines for physiological signals, including fNIRS, establishing foundations in signal representation, temporal structure, and robust analysis of noisy human-generated data.
This work helped shape a broader approach to sequential machine learning grounded in preprocessing, representation design, and evaluation under difficult real-world conditions.
Let’s connect
I am currently exploring Research Scientist roles — foundational or applied — as well as selective ML Engineering opportunities. If you are working on problems involving representation learning, sequential data, real-world robustness, or deployment-aware ML systems, I would be glad to connect.