Fateme Haghpanah
Download my resume!
I love connecting with nature through hiking, stargazing, and mountain climbing. Always eager to learn new skills and push my limits; whether exploring new ideas or lifting heavier weights with a barbell!
About me
I am Fateme, (/fɑ:te'me/), With dual Master’s degrees in Computer Science and Biomedical Engineering, I bring deep expertise in machine learning, honed through both academic and industry experiences. Recently, as a Machine Learning Scientist at Signal1, I developed a strong foundation in software development, backend coding, and tool building, particularly in Java. I excel in both independent and team settings, paying close attention to detail and proactively seeking impactful opportunities. I am committed to pushing boundaries and continuously expanding my skills.
Professional Experience
Machine Learning Scientist
Dec 2023 - August 2024
- Led the end-to-end optimization and migration of the CHARTWatch legacy machine learning pipeline from Kubeflow to a microservice architecture. Conducted in-depth data analysis, resolved critical issues, and implemented code fixes to align deterioration model scores with the main Java repository. Optimized model update frequency from 6 hours to 15 minutes, reducing maintenance overhead and enhancing system reliability and monitoring.
- Led an end-to-end machine learning-based Delirium Prediction feasibility study for a hospital client. Proactively learned about Delirium and researched its application to machine learning, despite no medical background. Designed and implemented a full ML pipeline to assess Delirium prediction feasibility using real-world hospital data. Evaluated model performance, explored explainability, and identified areas for optimization. Assessed the model's practical utility for early detection and its potential to reduce nursing staff workload in real-world hospital settings.
Graduate Research Assistant
Jan 2021 - May 2023
- Affiliated with, Vector Institute for Artificial Intelligence, University Health Network, and SickKids, the Hospital for Children.
- Designed and developed a novel classification algorithm to detect cerebral microbleeds (CMB) in preterm infants' MRI scans, achieving over 90% precision and recall. The neural network model is the first-ever developed one for infant brain scans.
- Developed a supervised convolutional neural network model to expedite the diagnosis of brain tumors in children within hospital emergency rooms with the AUC ROC score of 0.95.
Graduate Research Assistant
Sep 2018 - Sep 2020
- Implemented machine learning algorithms to segment distinct regions and tissues of neonatal brains, aiming to gain a better understanding of their development.
- Designed a novel neural network-based segmentation algorithm to enhance MRI safety by identifying various tissues of the entire body using different MRI scan modalities.
Data Scientist
Sep 2017 - Aug 2018
- Divar, is a classified-ad in Iran which is one of products of CafeBazaar.
- As the project owner, developed an innovative search algorithm that employs semantic networks and word embedding to enhance the relevancy of search results, ultimately leading to a better user experience.
- Utilized behavioral pattern analysis of unstructured data from user action logs to refine user queries and optimize search filters for various categories, resulting in an improved search experience.
- Developed a machine learning-based image search algorithm utilizing Convolutional Neural Networks to aid human reviewers in identifying and removing duplicate listings from the platform.
Publications
Published in Magnetic Resonance Imaging Journal, October 2022, Volume 92, Pages 140-149.
Authors: Xuzhe Zhang, Elsa D Angelini, Fateme S Haghpanah, Andrew F Laine, Yanping Sun, Grant T Hiura, Stephen M Dashnaw, Martin R Prince, Eric A Hoffman, Bharath Ambale-Venkatesh, Joao A Lima, Jim M Wild, Emlyn W Hughes, R Graham Barr, Wei Shen
Published in Brain Informatics Journal, December 2022, Volume 9, Pages 1-12
Authors: Yun Wang*, Fateme Sadat Haghpanah*, Xuzhe Zhang, Katie Santamaria, Gabriela Koch da Costa Aguiar Alves, Elizabeth Bruno, Natalie Aw, Alexis Maddocks, Cristiane S Duarte, Catherine Monk, Andrew Laine, Jonathan Posner, program collaborators for Environmental influences on Child Health Outcomes
Published in bioRxiv, May 2021
Authors: Yun Wang*, Fateme Sadat Haghpanah*, Natalie Aw, Andrew Laine, Jonathan Posner
Projects
- Implemented an encoder-decoder model with both single and multi-head attention mechanism to translate English to French using the Hansards dataset.
- Investigated the impact of causality on medical imaging by utilizing supervised machine learning algorithms to detect MRI acquisition parameters using the PPMI dataset, aiming to assess the potential improvement in performance.
- Developed a deep neural network for automatic identification of chest diseases from X-ray images using machine learning.
- Developed a Shiny App for data visualization to explore potential inquiries about the public dataset on housing litigation in NYC.
- Conducted a study on the lyrics of music from different genres using NLP techniques, including text preprocessing, data cleaning, topic modeling, and word embeddings like Word2vec.
- Created a facial emotion recognition classifier using feature extraction and selection techniques like HOG, SIFT, PCA, and landmark detection. Trained different classification models, including LDA, SVC, RandomForest, MLP, KNN, AdaBoost, and CNN-based approaches.
- Developed two collaborative filtering algorithms, Gradient Descent with Probabilistic Assumptions and Alternating Least Squares, from scratch. Subsequently, utilized Kernel Ridge Regression for post-processing the data.
- Developed models for landmark recognition using HOG features and advanced neural network based classification models.