Egor Marin

Machine Learning Scientist @ ENPICOM B.V., working with protein language models, bioinformatics, diffusion – you name it; @marinegor at most of the platforms.

PDF version of this CV is here

Profile

I have formal education in applied mathematics and physics (BSc + MSc), one year of full-time extracurricular education in computer science, PhD in biophysics and structural biology, and 8 years of computational lab experience.

I enjoy writing code, and want to write code that gets to run many times, and hence should be written wisely. I know a lot about (computational) biology, mostly on molecular level. And I can communicate with people.

Socials

• email: [email protected]
• github
• twitter
• linkedin
• bluesky
• google scholar
• ORCID

Career

I have been roughly 8 years in science, working with membrane proteins and their structure-function relationships: GPCRs, (microbal) rhodopsins, membrane transporters.

Have mainly worked at the Moscow Institute of Physics and Technology, and got my PhD from the University of Groningen. Also, I have worked at many synchrotrons and XFELs, and also was a visiting research assistant at the University of Southern California.

2024-current: Machine Learning Scientist

Doing machine learning in biotech-oriented SaaS company.

2017-2023: Scientist

• conducted research, managed data, wrote publications, participated in conferences
• managed students (BSs & MSc diploma), created a course on modern protein crystallography

2016-2017: Scientific Journalist

• analyzed publicational activity of MIPT
• wrote press-releases on published papers
• communicated with scientists & media.

Software skills & activities

Bag of words: python, numpy/sklearn/pytorch/lightning, polars🫶, huggingface🤗, uv/ruff❤️‍🔥, pytest, docker/compose, bash, mlflow, Ubuntu/nixos, HPC, SLURM.

💾 administrated ~15 Linux workstations and servers with ~40 users, managind ~200 Tb of research data.

😎 participated in Google Summer of Code contributing to MDAnalysis: introduced process-based parallelization to the library (see main PR).

🧑‍💻 contributed to opensource: reciprocalspaceship: wrote parser for serial crystallography data into binary dataframe-like class, ntfy-cryosparc: wrote web-server to parse CryoSPARC (tm) notifications and notify appropriate users.

🤓 MDAnalysis Core Developer since February 2025. For MDAnalysis, wrote a parallel backend for all analysis classes (dask/multiprocessing), added a DSSP module for native secondary structure assignment, currently working on an MMCIF parser.

🍝 performed large-scale calculations on SLURM and PBS, wrote bash scripts for reliable and reproducible data processing of serial crystallography data.

🏆 participated in data science competitions (top-10% in Kaggle “Predict Molecular Properties”, top-1 in first round of “Learning How To Smell”, top-10% in Takeda Signate competition, 5th place in Tochka Bank graph ML competition).

🤷‍♂️ self-hosted bunch of things: *arr, telegram bots, WebDAV, proxy & VPN servers, paperless, openwebui, you name it

🦀 Wrote a python(pyo3)+Rust(pest) parser for crystallographic data, contributed to polars-distance

Science skills & acitivities

Bag of words: structural biology, crystallography, cryoEM, cheminformatics, computer vision, data science, molecular docking, drug discovery, protein structure, GPCRs, membrane proteins, structure-based drug discovery, antibodies, protein language models, discrete diffusion, flow matching, AlphaOpenfold

🧬+🥩 structural biology: co-published papers in Science, Nature Communications, JACS, Science Advances, Journal of Chemical Information and Modelling, Scientific Data. Performed data analysis, wrote texts, created figures, managed writing process – the normal stuff.

💊 structure-based drug discovery: performed large-scale virtual screening campaign, created robust accelerated virtual screening approach, communicated with CROs, oversaw functional tests.

🤖 machine learning: did many ad-hoc ML applications in computer vision (background removal with NMF decomposition), clustering, supervised learning. Always try linear regression first, have a paper about it.

👾 deep learning: know about protein language models and their properties, AlphaOpenfold-like models and their applications, wrote toy discrete diffusion models, adapted open-source discrete diffusion models for other domains.

Education

Moscow Institute of Physics and Technology, 2013-2017 BSc in applied mathematics and physics, magna cum laude

Moscow Institute of Physics and Technology, 2017-2019 MSc in applied mathematics and physics, summa cum laude, with specialization in biophysics and structural biology

Computer Science Center, 2020-2022 Full-time extracurricular education in computer science: Python, C++, Algorithms and Data Structures, Data Science, Intro to Linux Systems, Rust

University of Groningen, 2019-2023 PhD, diploma on “On the methods of studying protein-ligand interaction dynamics”

Last updated: May 2025.