Pin-Wei Benny Chen, PhD

👋

Machine Learning & Wearable Tech for Clinical Research

I work at the intersection of machine learning, wearable technology, and clinical research — developing algorithms and data pipelines that turn continuous sensor data into clinically meaningful insights. Over 10 years, I've led ML development and large-scale clinical trials at institutions such as Children's Hospital and Rehabilitation Center. My work spans actigraphy, sleep staging, activity recognition, and remote monitoring for research teams and medical device companies. I hold a PhD in Rehabilitation Science and a Master in Mind, Brain and Behavior. I have supported 7 NIH R01 grants and over $13M in funded research.

Get in Touch View CV

Services

( 01 )

ML Model Development

A validated predictive model or classifier trained on your clinical or wearable dataset, ready for publication or deployment

I build and validate machine learning algorithms for clinical datasets and longitudinal time series data using rigorous MLOps practices. My end-to-end development cycle ensures reproducibility from data prep to deployment. Deliverables include documented model performance metrics, validation protocols, pilot studies suitable for clinical publication and grant application, and production-ready code in R and Python." Suited for research teams with labeled data who need a defensible, publication-ready ML component, or medical device companies requiring algorithm development for FDA submission or grant deliverables.

Remote Monitoring Consultation

A technically sound algorithm development and clinical validation strategy — so your device meets regulatory and publication standards

I advise medical device and wearable technology companies on the full arc from algorithm concept to clinical validation: study design, sensor data collection protocols, ML algorithm development, and performance benchmarking against clinical reference standards. Suited for early-stage medical device companies preparing grants or FDA submissions, and research teams integrating wearables (e.g., Apple Watch, Fitbit, Actigraph) as well as other remote monitoring sensors (e.g., blood sugar) into IRB-approved clinical protocols.

Data Pipelines

A clean, reproducible processing pipeline for your data — so your research team spends time on analysis, not data wrangling

I design and implement cloud-based data science pipelines for high-resolution sensor data: raw signal ingestion, preprocessing, feature extraction, and output formatting for downstream analysis. Validated against clinical reference standards and built for multi-site research deployment. Webapp development for output presentation. Suited for teams running longitudinal cohort studies with remote monitoring data who need scalable, automated processing with audit-ready documentation.

Highlighted Experience

( 02 )

Apr 2024 – Present

Mobile Health Data Scientist

Children's Hospital of Philadelphia

Design scalable cloud-based data pipelines for sleep and activity metrics. Develop ML algorithms for human activity detection. Support 3 NIH R01 grant applications across 10+ projects.

Jun 2021 – Mar 2023

Project Lead

Shirley Ryan AbilityLab

Led ML algorithm development for sleep stage detection in stroke populations. Managed a 9-member team through a large inpatient clinical trial. Secured a $3M DoD grant for wearable sensor research and delivered a $4M government contract.

Nov 2018 – Oct 2019

Co-founder & Director

Proprio / PlatformSTL Incubator

Secured a $100k government grant for an ML-based digital health system for stroke monitoring. Led a cross-functional team of engineers, scientists, and physicians to build an MVP medical software product.

Aug 2017 – Aug 2018

Chief Communication Officer

Sling Health (Non-profit Incubator)

Negotiated university partnerships saving $20k in marketing. Secured key investment partners to expand startup teams and accelerate growth.

Selected Projects

( 03 )

Modular Actigraphy Platform at Children's Hospital of Philadelphia

Processing support for 10+ active research projects

Children's Hospital of Philadelphia

Challenge

Clinical researchers lacked a scalable, automated pipeline for processing raw wearable sensor data (actigraphy) for sleep and physical activity assessment in pediatric populations. Manual processing was creating a bottleneck across multiple concurrent studies. Proprietary softwares are blackboxes that doesn't feed the research needs.

Solution

Built a Modular Actigraphy Platform — a cloud-based data science solution for processing high-resolution time series sensor data. Include scientific validated algorithms reference standards.

Outcome

Platform actively supports 10+ concurrent research projects. Work contributed to numerous NIH R01 grant applications. Results are available as a preprint on medRxiv.

actigraphy

sleep

pediatrics

cloud-pipeline

NIH

Apple Watch Activity Recognition for Post-Stroke Remote Monitoring

Working MVP delivered; $100k SBIR grant secured; 1 peer-reviewed publication

Proprio

Challenge

Post-stroke patients had no way for clinicians to monitor daily activities remotely and objectively. Existing solutions required clinical-grade hardware or were not accurate nor convenient.

Solution

Led algorithm development for an Apple Watch-based activity recognition system for post-stroke patients. Collaborated across engineering, science, and physician teams. Secured a $100k SBIR grant to build and validate the MVP from concept through clinical pilot.

Outcome

Working MVP delivered and validated. Results published in *IJERPH* (2021), Chen et al.

activity-recognition

stroke

Apple-Watch

SBIR

remote-monitoring

Automated Sleep Monitoring in Acute Stroke Rehabilitation

Delivered under $4M government contract

Shirley Ryan AbilityLab

Challenge

No accurate algorithms existed for sleep monitoring in acute stroke rehabilitation. Manual observation was impractical at scale, leaving clinical researchers without objective sleep data in inpatient settings.

Solution

Designed and implemented an ML algorithm for sleep stage detection using multimodal wireless sensors. The development acocompany a large in-patient RCT trials with a team of 10-person interdisciplinary team where I led the effort in the RCT trial and the development of the algorithms.

Outcome

Results delivered under a $4M government contract. Approach published in *Sensors* (2022), Chen et al.

sleep

stroke

wearables

clinical-trial

multimodal-sensors