Jiarui Yu

Seattle,Washington

Overview

years of professional experience

National Oceanic And Atmospheric Administration, NOAA

06.2023 - 09.2023

Investigated coastal upwelling dynamics using machine learning algorithms on remote-sensing data, contributing to the understanding of global oceanic patterns.
Developed predictive models for coastal upwelling intensity using neural networks and state-of-art models (Transformer, ConvLSTM, U-Net, etc.)
Demonstrated flexibility and adaptability in response to project needs by developing and implementing custom deep learning solutions for ocean remote-sensing data.

University Of Washington

02.2023 - Current

Set up energy measurement toolchains (both hardware setup and software apps) to support collecting end-to-end energy consumption ground truth data using the hardware power monitor.
Built a predictive energy estimator by assisting in benchmarking selected DNN models and applications and collecting energy-related features.
Built software apps for both iOS and Android to enable users to interact with deep learning CNN models

NSTEM

08.2022 - 12.2022

Simon Fraser University

Burnaby, BC

12.2025

University of Washington

Seattle, WA

03.2024

Programming Languages: Python, MySQL, Java, R, Javascript, HTML, CSS, MATLAB, Swift
Machine Learning: TensorFlow, PyTorch, Keras, sklearn
Software Development: Xcode, Android Studio, Streamlit
Data Analysis: Pandas, Numpy, GeoPandas
Visualization: Matplotlib, Seaborn, Plotly

Cloud Service Platforms: AWS, Google Cloud, Microsoft Azure
Hardware: Monsoon
Optimization: Pulp, Gurobipy
Programming Skills: Dynamic Programming, Divide and Conquer, backtracking, Union Find, Topological Sort, Data Stream, Monotonic Stack, Trie
Certificates: HackerRank Python Certificate, HackerRank SQL Certificate, HackerRank Javascript Certificate

Led a team at OceanHackWeek (University of Washington) to develop deep learning models, including Transformers, ConvLSTM, and 3D CNN, for time-series data prediction, enhancing predictive accuracy and model performance.
Engineered and optimized image classification models using CNN architecture to accurately identify coastal upwelling, and devised predictive multivariable models combining ConvLSTM and Transformers for Sea Surface Temperature Data prediction, incorporating LIME for interpretability and U-Net for detailed analyses.
Created a full-stack application using Java and SQL to manage interactions among patients, caregivers, and vaccines, supporting functionalities like appointment scheduling, availability checks, and registration processes.
Developed a full-stack web platform for online food shopping, employing Javascript, HTML, CSS, Node.js, and SQL, enabling customers to make purchases and provide feedback, with a system to track purchase history and enhance user experience.
Applied linear and non-linear optimization techniques in Python to construct an optimized scheduling system for a Seattle-based restaurant chain, balancing the demands of employees and employers, and gaining recognition through social service credit.