Hey hey! I’m Modestas, a MSc student at ETH Zürich, Switzerland in the Departments of Computer Science and Biosystems. Lately I’m focusing on Natural Language Understanding applications, especially in Biomedical and Healthcare domains. In college I studied Chemistry, but I have just enough expertise in computer science and machine learning to make trouble. My research and broader service are all grounded in the idea that science and technology are inherently social activities, which are directed, operated, and maintained in certain ways by people in and between institutions. Thus I’m curious to explore how the latest technological advances in language understanding can help disentangle the patient-healthcare institution relationships, or even elucidate the most abstract evolutionary relationships such as protein function by treating protein sequences as words.

My work and research

Currently I’m writing my master thesis at IBM Research, Zürich in pretraining self-supervised language models for protein classification tasks. Self-supervised learning established itself as a powerful method for learning useful information from unlabeled sequences. In this master thesis project, I’m utilizing the latest NLP embedding techniques to learn a protein language model, and fine-tune it on three types of protein classification tasks. We’re making an assumption here, that amino acids can be treated as words, which make up proteins - the sentences of Biology.

My background and history

I received my BSc from Macalester College in Chemistry. Afterwards, I spent some time at University of Washington in Seattle researching HIV Envelope protein biophysics, followed by a move to Basel, Switzerland to work as a Bioinformatician at a cancer neo-antigen discovery startup Specifica Inc.

During my Bioinformatics masters studies, I have worked at Novartis Institutes of Biomedical Research on multimodal learning to generate doctors’ notes from chest X-ray images.