Tony Passaro, PhD
Building the data and modeling systems that power
biosignal-based products
From raw, noisy biosignals to production-ready systems that are built to work in real-world conditions, including real-time and predictive applications. 0+ years across signal processing, ML, and deployment.
Most teams can build models in controlled environments. Very few can get those models to hold up in real-world data and be ready for deployment.
That's where I come in—turning noisy biosignal data into models that are stable, validated, and ready to be integrated into production systems.
What I Do
Core Capabilities
Signal Processing
- Large-scale multimodal time-series analysis (EEG, ECG/HRV, respiration, GSR, EMG, EOG)
- Artifact rejection and signal extraction in high-noise, real-world environments
- Spectral, temporal, and connectivity-based feature engineering
- Robust preprocessing pipelines for heterogeneous, real-world datasets
Machine Learning & Modeling
- Predictive modeling of physiological and behavioral states
- Time-series modeling and high-dimensional feature systems
- Model development focused on generalization beyond controlled environments
- Validation, calibration, and stabilization of models for real-world performance
Deployment-Ready Modeling
- Designing models and pipelines ready for integration into production systems
- Real-time and forward-prediction modeling under practical constraints
- Optimization for latency, stability, and reliability in deployment settings
- Close collaboration with engineering teams for system integration
Expertise
Modalities & Data Types
EEGECG / HRVGSR / EDARespirationEMG / EOGEye TrackingPupillometryAccelerometryMEGfMRIDTI
Portfolio
Selected Work
Selected examples of building models that operate reliably in real-world biosignal systems
Attune
Focused Ultrasound Sleep Wearable
- Built a single-channel EEG sleep staging model (~90% accuracy) with forward prediction, enabling closed-loop neuromodulation timing that would not be possible with standard retrospective staging
- Implemented spindle and slow-wave detection for real-time, phase-aware stimulation, allowing precise alignment of stimulation with endogenous neural events
- Developed subject-specific slow-wave modeling, significantly improving signal fidelity and reducing variability across noisy, heterogeneous datasets
- Designed peak-frequency-based control models for FUS modulation, enabling dynamic sleep induction and awakening within constrained stimulation windows
- Built artifact rejection methods based on spectral ratios, improving model robustness across real-world, multi-site EEG recordings
- Contributed to DARPA-funded programs, transitioning models from research-grade performance to deployment-ready neuromodulation frameworks
Breethly
Respiratory Wearable Platform
- Built real-time breath segmentation models from raw respiratory signals, enabling structured interpretation of breathing behavior in a production wearable environment
- Developed scoring models for breathing accuracy, enabling quantifiable feedback and calibration during guided breathing exercises
- Designed feature pipelines for physiological metrics (lung capacity proxies, AUC, breath dynamics), allowing translation of raw signals into meaningful physiological outputs
- Engineered models robust to motion artifacts and real-world variability, enabling reliable performance outside controlled conditions
- Translated raw signals into real-time feedback systems, enabling user-facing behavioral guidance driven directly by physiological data
NeuroLight
EEG Audio-visual Sleep Mask
- Built EEG-driven models to induce and stabilize sleep states, enabling closed-loop sensory modulation based on live neural activity
- Identified spectral and temporal EEG features linked to sleep quality, enabling data-driven optimization of stimulation protocols
- Developed closed-loop modeling frameworks, allowing continuous adaptation of stimulation based on real-time neural signals
- Contributed to identification of transferable neural signatures, supporting generalization across users rather than subject-specific tuning alone
Eno
EEG Wearable Headphones
- Built EEG feature extraction and classification models, enabling real-time detection of cognitive states in a low-channel consumer device
- Developed models for focus, relaxation, and cognitive transitions, enabling usable state feedback despite limited signal quality
- Implemented neurofeedback and entrainment models under hardware constraints, allowing functional performance within tight device limitations
- Solved low-channel, high-noise signal challenges, enabling practical usability of EEG in real-world consumer environments
ZeitMedical
Clinical Stroke Detection
- Built EEG-based stroke detection models, enabling rapid, pre-hospital identification of stroke risk from noisy clinical data
- Developed models robust to extreme noise and variability, enabling operation in uncontrolled emergency and field conditions
- Led interdisciplinary team (clinicians + postdocs), accelerating development and alignment between clinical and modeling requirements
- Contributed to clinical validation and integration strategy, supporting translation from model development to real-world clinical deployment
U.S. Army Research Laboratory / DCS Corporation
Neural Performance Research
- Led efforts to extract reliable signal from large, noisy, real-world neural datasets, enabling identification of features that persist outside controlled environments
- Developed predictive models linking neural activity to real-world performance, enabling quantification of cognitive and behavioral outcomes from biosignals
- Designed and validated feature extraction and modeling pipelines, establishing repeatable methods for large-scale EEG and multimodal analysis
- Identified robust, generalizable features, enabling transferability across subjects and conditions rather than overfitting to lab data
- Built tools and frameworks for artifact detection and large-scale analysis, enabling efficient processing of high-volume, real-world datasets
Research
Google ScholarAcademic Foundation
Background spanning clinical and cognitive neuroscience, with deep experience in EEG, MEG, and fMRI across patient and normative populations.
Contributed to 0+ peer-reviewed publications involving epilepsy, stroke, Alzheimer's, ADHD, and other neurological conditions.
Funding
Grants & Programs
Work supported by:
DARPA Neurotechnology Programs
SBIR / STTR Programs
U.S. Army Research Laboratory
AFRL & SOCOM Initiatives
Department of Defense Research
Government-funded Lab Development
Engagement
How I Operate
Contract / advisory / part-time engagements
End-to-end ownership of data and modeling
Operate independently or within small teams
Rapid development of production-ready systems
Raw data to deployed system without handoffs
Typically brought in to solve the most technically uncertain part of a product and move it to a working, deployable model quickly.
● Available Immediately
Let's Build Something
Have a biosignal system that needs to work in production?
Get in Touch