Cleah Winston

B.S./M.S. in Computer Science @ University of Washington

Robotics Research Student @ UW Robotics

Teaching Assistant @ Paul Allen School, UW

About Me

I'm an undergraduate at the University of Washington, Seattle, pursuing a B.S./M.S. in Computer Science with a minor in Business Administration. I conduct research on reinforcement learning and robotics, advised by Abhishek Gupta. I'm interested in making robot policies that are sample efficient and use data effectively, whether it be through utilization of non-expert data or through combining offline and online data. I've also explored using large language models to enhance the clinical workflow and patient care. Broadly, I'm passionate about using AI to drive real-world impact and improve the efficiency of complex systems.

Publications

RISE Project

Using Non-Expert Data to Robustify Imitation Learning via Offline Reinforcement Learning

In Review, ICRA 2025

Kevin Huang*, Rosario Scalise*, Cleah Winston, et al. [show all]

Paper Website
Medical Question-Generation for Pre-Consultation

Medical Question-Generation for Pre-Consultation with LLM In-Context Learning

GenAI4Health, NeurIPS 2024 Poster

Caleb Winston, Cleah Winston, et al. [show all]

Paper
Multimodal Clinical Prediction

Multimodal Clinical Prediction with Unified Prompts and Pretrained Large-Language Models

IEEE International Conference on Healthcare Informatics, 2024

Caleb Winston, Chloe Winston, Cailin Winston, et al. [show all]

Paper
Repairing Brain-Computer Interfaces

Repairing Brain-Computer Interfaces with Fault-Based Data Acquisition

IEEE International Conference on Software Engineering, 2022

Cailin Winston, Caleb Winston, Chloe N Winston, et al. [show all]

Paper
Precisely Detecting and Resolving BCI Errors

Precisely Detecting and Resolving BCI Errors With Case Functions

IEEE Engineering in Medicine & Biology Society, 2021

Chloe N Winston, Cailin Winston, Caleb Winston, et al. [show all]

Paper

Projects

Using Non-Expert Data to Robustify Imitation Learning with Offline RL

September 2024 - October 2025

Improved performance of imitation learning on non-expert data using offline reinforcement learning. By enforcing Lipschitz continuity on the behavior policy and utilizing data augmentation to widen the action space of the data, stitching between expert and non-expert states was achieved.

WEIRD Lab @ UW In Review, ICRA 2025 Website

LLM-based Pain Logging App

January 2025 - Present

Developed a pain logging app using Gemini models to process user inputs, display structured pain events, and visualize pain metrics in an intuitive UI. This platform allows users to simply enter in any event (pain levels, medication intake, diet, etc.) and the LLM will categorize events.

UW, Stanford University, University of Pennslyvania App

Large Language Models for Clinical Predictions

October 2023 – September 2024

Developed a chatbot that collects patient information to curate a history of present illness (HPI).

UW, Stanford University, University of Pennsylvania 2024 International Conference of Health Informatics, 2024 GenAI4Health (NeurIPS) 2nd @ MD Plus Datathon on value-based care

Improving Robotic Car Navigation Algorithms

September 2022 – May 2024

Trained convolutional neural networks to predict steering angles to follow a road from simulated data from AirSim. Investigated whether robust training can be achieved with simulated data and less real data, collected from MuSHR cars.

Robot Learning Lab @ UW 2024 Undergraduate Research Symposium

Augmented Reality for Physical Therapy

September 2023 – February 2024

Developed a prototype for an AR-based system using the Meta Quest Pro that creates a more immersive and interactive form of doing at-home physical therapy. Created a business plan for the product and pitched at the 2024 Holloman Health Innovation Challenge, UW.

UW CSE, Foster School of Business, UW Medicine Finalist at Holloman Health Innovation Challenge

Repairing Brain-Computer Interfaces (BCI) with Fault-Based Data Acquisition

April 2021 – May 2022

Developed a method to improve the accuracy of brain-computer interfaces (BCIs) using fault-based data acquisition. This approach identifies when the BCI is making errors and collects additional training data to correct them.

UW CSE 2022 International Conference on Software Engineering

Distracted Driver Detection Using EEG

January 2021 – March 2021

Designed an ML-based system to predict distracted driving from EEG data. Collected data through a user study where subjects were placed in distracting situations while operating a virtual car. Trained regression models and neural network decoders.

UW CSE

MyScoliCare: An App for Scoliosis Care

November 2018 – May 2019

Developed an Android and iOS app for motivating scoliosis patients to complete Schroth therapy and daily bracing. The app was endorsed by Dr. Manuel Rigo (MD/PhD), the head of the Barcelona Scoliosis Physical Therapy School, and has reached over 100 therapists globally and many scoliosis patients.

Conference on Non-Operative Treatment of Childhood Scoliosis, Columbia University More info

PriceNet

November 2018 – May 2019

Designed a fruit ripeness sensor and a reinforcement-learning system that processes sales data and ripeness sensor data to suggest effective selling prices to minimize fruit wastage. Created a circuit system that detects the color of fruits based on photoresistor signals.

1st place at WSSEF 2019, Naval Science Award, Broadcom MASTERS nominee

Investigating Cybersecurity in Driverless Cars

September 2017 – March 2018

Built a robotic car by programming an Arduino UNO board to understand the risk of cyberattacks on driverless cars. Designed and conducted experiments to investigate how cybersecurity is a real issue for autonomous vehicles.

1st place at WSSEF 2018, top 300 Broadcom MASTERS nominee

Work Experience

Supio

Software Engineering Intern | June 2025 - Aug. 2025

I contributed to developing a platform that allows internal users to easily fine-tune LLMs. Using AWS S3 for data storage and AWS EC2 instances for training, we abstracted away the complications of data generation, training, and evaluation.

Paul Allen School of Computer Science

Teaching Assistant | Sept. 2023 - Present

I am a 5th time teaching assistant (TA) for machine learning, discrete math, and probability classes for computer science. As part of being a TA, I lead weekly ~30 student sections, assist students through office hours and 1-1 sessions, help with course logistics (e.g., running the website), host review sessions, and proctor exams.

Neurophys

Software Developer | Aug. 2024 - Sept. 2025

I developed and trained ML models that detect stroke in patient voice recordings with limited data.

Banyan Computing

Backend Developer Intern | Sept. 2021 – April 2022

At Banyan, I developed Banyan Processes, a feature that enables users to save Julia scripts to the cloud and schedule them for later execution. Additionally, I implemented features to allow users to manage their cloud spending.

Skills

Languages

Python Java HTML/CSS JavaScript React JS Julia C SQL

Libraries

PyTorch Scikit-learn NumPy Pandas AWS

Technical Skills

Machine Learning Website Development Data Analysis Experiment Design