Ayse Berceste Dincer

I am a Senior Software Engineer at Uber Technologies, working with the Marketplace team. My current role encompasses software engineering, machine learning engineering, and research in applied ML. I focus on developing machine learning models for improving the marketplace efficiency as well as managing the entire ML model workflow including experimentation, productionization, and maintenance of models/pipelines.

I received my Ph.D. in Computer Science from the University of Washington, Paul G. Allen School of Computer Science & Engineering in 2022 where I was advised by Prof. William Stafford Noble and Prof. Su-In Lee. My Ph.D. research focused on developing and using machine learning techniques to solve biomedical problems, including representation learning for high-dimensional data and integration of denoising methods into deep learning models.

Contact

bercestedincer at gmail.com

My CV

Selected Projects and Publications

Multitask neural network-based driver incentive generation | Uber Technologies

Developed deep multitask neural network models to provide drivers with customized incentives which significantly improved the overall incentive efficiency and drove the project across different stages including ideation, model formulation, development and iterations, experimentation, analysis, launch, and real-life maintenance of models/data pipelines/workflows.

Project1

A deep learning approach to eliminate bias in protein quantification | Noble Lab, University of Washington

Developed a convolutional neural network (CNN) model to predict multiplicative noise coefficients from sequences which reduced protein quantification noise and outperformed alternative models.
Dincer, A. B., Lu, Y. Y., Schweppe, D. K., Oh, S. & Noble, W. S. (2022). Reducing Peptide Sequence Bias in Quantitative Mass Spectrometry Data with Machine Learning. J Proteome Res., 21(7), 1771-1782. (Paper) | Contributed talk at ASMS 2021 and ISMB/ECCB 2021 (best presentation award).

Project2

Adversarial Deconfounding Autoencoder for learning robust embeddings | AIMS Lab, University of Washington

Developed an unsupervised deep learning approach for learning deconfounded embeddings and improved cancer subtype classification across different data domains.
Dincer, A. B., Janizek, J. D., & Lee, S. I. (2020). Adversarial Deconfounding Autoencoder for learning robust gene expression embeddings. Bioinformatics, 36(Supplement 2), i573–i582. (Paper) | Contributed talk at ISMB MLCSB 2020. (Talk)

Project3

DeepProfile: Interpretable deep learning of latent variables for 18 human cancers | AIMS Lab, University of Washington

Collected and integrated gene expression measurements from 1,098 datasets and 18 cancer types and increased the robustness of variational autoencoders (VAEs) by designing an ensemble learning pipeline.
Qiu, W., Dincer, A. B., Janizek, J. D., Celik, S., Pittet, M., Naxerova, K., & Lee, S. I. (2023). A deep profile of gene expression across 18 human cancers. Under review in Nature Biomedical Engineering | Spotlight talk at Machine Learning in Computational Biology (MLCB) 2019. (Paper)

Project4