Arthur Qin

Performing Deep-Learning Streamflow Forecasting of the Russian River Basin in collaboration with the Environmental Systems & Dynamic Laboratory by building and adapting LSTM models (via NeuralHydrology) using observational, physically based, and CAMELS datasets. Experimenting with diverse architectures (e.g., multi-scale LSTM), incorporating physical constraints, and parallelizing hyperparameter tuning on high-performance computing clusters.

Coaching Lincoln Douglas debaters with evidence and strategy preparation, and judging Lincoln-Douglas debates.

Led independent research project for the Kaggle '25 Brain-to-text competition, optimizing an RNN + 5gram LM predicting speech output from iEEG neural activity.

Translated offline MATLAB MINT neural decoder algorithm into a streaming architecture prepared for clinical deployment, implemented through a clean, modular Python framework that enhanced performance, scalability, and real-time integration. Generalized its implementation for other datasets, and integrated it into ezmsg signal processing pipeline.

Under Professor Laurel Larsen, performing Low-Latitude Peatland Mapping with ML by consolidating and comparing existing peatland maps, then build labeled training sets using peat vs. no-peat accumulation. Gathering geospatial layers to inform peat accumulation and developing a machine-learning classifier to poorly documented regions for expanded peatland insights.

Research and evidence preparation for nationally ranked debate team and judging policy debates, providing feedback and strategic insights to enhance competitors’ performance. Extensive experience in high-level research, argumentation, and leadership through mentorship of younger debaters.

Research internship at LMU under Professor Lei Huang. Tasked to build machine learning pipeline that incorporated over a dozen spatiotemporal variables aggregated from Google Earth satellite data. Trained and evaluated models to predict the spread of active wildfires - independently developed a cloud-based training pipeline, leveraging rented GPU pods and GitHub for efficient iteration and deployment. Received an award from NOAA for innovative environmental research through its “Taking the Pulse of the Planet” program.

Worked onsite with mentor on a Web API to access and transfer COVID-19 test result data for over 200k patients from Fulgent laboratories and efficiently upload to long term cloud-based storage. Involved transfers from an Amazon S3 Bucket to DynamoDB to archive and process CSV data in Parquet format. Involved using Cloud Automaton with Amazon EC2 to enable cloud data transfers and save local processing power and time. Worked with a light Python web framework, including Flask and AWS S3.