Sarah H. Amiraslani

self.png

Hello! I’m Sarah, a data professional with a passion for learning and a desire to solve challenging problems.

I have 5+ years of experience working in data analytics across diverse industries including manufacturing automation, quantitative modeling, AI model development, and academic research. I am currently working on building data engineering pipelines to support NVIDIA NeMo products for generative AI.

I’m passionate about learning from others and mentoring. If you’re interested in exchanging ideas or exploring potential collaborations, please feel free to reach out.

Selected projects

  • Predicting Sunspot Activity with Machine Learning

    Predicting Sunspot Activity with Machine Learning

    This project leverages ensemble and deep learning models to forecast sunspots and Heliospheric Current Sheet (HCS) indexes, enhancing the prediction of solar wind structures.

    Technologies used: ARIMA, Seasonal naïve, LTSM, Prophet, AR-Net, Extra Tree

  • Tracing the Origins of Solar Wind with Machine Learning

    Tracing the Origins of Solar Wind with Machine Learning

    This project uses in-situ measurements and unsupervised learning to cluster solar wind with similar properties and map it to its coronal origins, aiming to enhance predictions of heliospheric phenomena through machine learning.

    Technologies used: Unsupervised Learning, PCA, t-SNE, DBSCAN

  • Predicting Top 10 Formula 1 Finishes and Clustering Race Tracks

    Predicting Top 10 Formula 1 Finishes and Clustering Race Tracks

    This project applies supervised learning to predict top 10 race finishes, pinpointing key factors that influence these outcomes. Additionally, it employs unsupervised learning to cluster race tracks based on characteristics like layout and surface, offering novel insights into racing strategies and optimizing performance.

    Technologies used: Random Forest, Logistic Regression, Neural Networks, DBSCAN

  • Draw to Learn: A Strategy to Learn Abstract Concepts

    Draw to Learn: A Strategy to Learn Abstract Concepts

    This project examines the effectiveness of student-generated drawings in learning abstract science concepts. By comparing various visualization strategies and important covariates in a lab setting, we explore how different methods impact retention and understanding of complex topics like black holes.

    Technologies used: Experimental Design, Multivariate Analysis, General Linear Models

Latest blog posts

Jul 22, 2024 Exploring Breaking Bad Relations
Jun 28, 2024 Unraveling Complexity: Network Science Applied to Star Wars Films