Active
Music Perception via Physiological Signals
Decoding musical factors such as genre, tempo, mode, and register directly from EEG and EDA signals.
Hi there!
I'm Itir Sayar
A PhD student in Computer Science at UMass Amherst, where I also completed my undergrad with BS degrees in CS and Neuroscience.
I build closed-loop wearable systems with neurophysiological sensing for everyday health and wellness.
My research spans designing unobtrusive sensing systems, running user studies for human data collection in daily-life scenarios, and building machine-learning models that close the loop between those signals and the people.
What I work on
01 · SENSE
Sense
Designing ubiquitous systems that capture physiological signals from people in everyday, real world scenarios.
EEGEDAEye tracking
02 · MODEL
Model
Developing machine-learning models, mainly self-supervised and multimodal methods, that turn physiological streams into reliable representations.
Self-supervisedMultimodalFoundation models
03 · UNDERSTAND
Understand
Connecting those representations to human cognition, emotion, and behavior, then closing the loop with systems that adapt to people.
CognitionAffectHCI
Research
/ selectedPhD · UMass Amherst
Active
Multimodal Physiological Sensing for Engagement
A deep-learning model fusing EEG, EDA, and eye tracking to model viewer engagement during educational content, treating engagement as a rich multimodal signal rather than a single self-reported number.
Undergraduate
2023 – '24
Lymphoma Survival Prediction via Radiomics
Honors thesis (Biomedical Imaging & Data Science Lab). An ML/radiomics pipeline for progression-free survival in DLBCL from 3D PET imaging. Submitted to the SNMMI AI 2023 Challenge.
Summer '23
Biologically-Inspired RL with TD3
At the BINDS Lab. Combined muscle memory with TD3 + dynamic gating, cutting computational cost ~60% while holding performance across long action sequences.
Poster · CICS First Friday Fair
2023
App Danger Project — Child Online Safety
With Prof. Brian Levine's UMass Rescue Lab. AI data labeling for a large-scale study of exploitation indicators across app-store reviews, supporting app-danger.org. Covered by The New York Times and The Wall Street Journal.
Teaching & Mentorship
Teaching
COMPSCI 345 · Data Management
Weekly office hours, lab support for SQL and relational design, progress tracking across assignments and exams.
COMPSCI 365 · Digital Forensics
Course development: lab exercises, forensics assignments, and grading rubrics.
COMPSCI 563 · Internet Law & Policy
Led discussion sections on internet law and digital policy; graded policy analyses.
COMPSCI 345 · Data Management
Guided 197 students through SQL and database concepts; authored exam questions, proctored exams.
Mentorship & Service
Research Mentor · ERSP
Mentoring undergraduate researchers through the Early Research Scholars Program on personalized health insights & wearable sensing.
CRA-WP Grad Cohort
Selected participant.