Şafak Bilici

This is my personal blog. I mostly post about mathematics, computer science, linguistics, and music.

About

Şafak Bilici

How to pronounce my name: (sha-FUCK billie-G) or IPA -> ([ʃɑ.ˈfɑk] [biɭiʤi])

I am an AWS certified Machine Learning Engineer with robust experience in large-scale distributed backend systems, cloud services, search engines, and data science in general. My expertise lies in solving real-world, high-scale problems related to low-latency information retrieval, personalization, and product discovery using modern machine learning and software engineering technologies. Currently, I am focused on tackling challenges such as cross-lingual information retrieval, multi-modal search, and leveraging large language models for search.

I am currently working at Insider’s Griffin Team as a Machine Learning Engineer, wrangling complex search applications using state-of-the-art machine learning and software engineering techniques. As Griffin, we aim to personalize search results for online shoppers with AI-led recommendations. Before joining Insider, I was a Research Engineer at Huawei’s AppGallery Search team. I developed AppGallery’s search features such as semantic search, spelling correction, query expansion, etc. for more than 50 million users in Russia, Arab countries, Latin America, and Asia countries excl. China. Before Huawei, I was an ML researcher at YTU Nova Research Lab. under the supervision of Prof. Dr. Fatih Amasyali. At Nova Lab., I mostly worked on the intersection of variational inference and low-resource language models.

I did my BSc @YTU CS. My BSc thesis was on multimodal transformers and natural language grounding, with applications on low-resource cross-modal retrieval, visual question answering, and few-shot image classification.

My research interests are information retrieval, machine translation, and Bayesian machine learning in general. Besides that; I love population biology, language acquisition, and biolinguistics. To express my field with non-technical terms: I teach computers to understand human language, and I (try to) approximate difficult-to-compute probability densities.

There are things that I like much more than science: music. I have been studying classical music theory for almost 10 years. If something comes to my mind, then I compose. My main instrument is piano, however, I was once a guitar player and vocalist in a death metal band.

If you want to discuss the things that I mentioned above or anything else, buy me a coffee. My email address is at the bottom of the page. Damn, I really love coffee.

Publications and Preprints

Can Özbey, Talha Çolakoğlu, M. Şafak Bilici, Ekin Can Erkuş, “A Unified Formulation for the Frequency Distribution of Word Frequencies using the Inverse Zipf’s Law”, in Special Interest Group on Information Retrieval (SIGIR), 2023. (paper)

M. Şafak Bilici, Mehmet Fatih Amasyali, “Transformers as Neural Augmentors: Class Conditional Sentence Generation via Variational Bayes”, arXiv preprint arXiv:2205.09391, 2022. (paper, repository)

E. Sadi Uysal, M. Şafak Bilici, B. Selin Zaza, M. Yiğit Özgenç, Onur Boyar, “Exploring The Limits Of Data Augmentation For Retinal Vessel Segmentation”, arXiv preprint arXiv:2105.09365, 2021. (paper, repository)

M. Şafak Bilici, Mehmet Fatih Amasyali, “Variational Sentence Augmentation For Masked Language Modeling”, in Innovations in Intelligent Systems and Applications Conference (ASYU), 2021. (paper, repository)

Software

bayesmedaug: bayesmedaug optimizes your data augmentation hyperparameters for medical image segmentation tasks by using Bayesian Optimization and Gaussian Process.
x-tagger: x-tagger is a Natural Language Processing toolkit for sequence labeling in its simplest form.