maiia-kotyga-network@yandex.ru
Hi đź‘‹
I’m Maiia Kotyga, a data analyst and AI specialist focused on machine learning, mathematical modeling, NLP, and healthcare digitalization.
My work combines analytical and ML-based solutions, data infrastructure, applied research, and educational projects. I have a background in mathematics and IT, and experience in product, technology, and research contexts.
My current research interests include efficient LLM inference, with a particular focus on dynamic context-dependent pruning methods.
I teach and mentor in AI, NLP, and data science, support hackathon and olympiad teams, and create educational materials for students and early-career ML practitioners.
In short: a human teacher for machine learning 🤖.
Worked on healthcare digitalization projects, focusing on data-driven decision-making, forecasting, analytics, and database architecture.
Taught and mentored students at the Deep Learning School by the Moscow Institute of Physics and Technology.
Visiting lecturer for the “Digital Tools for Financial Analysis” discipline within the “Corporate Financial Analyst” program.
Mentored student teams during a hackathon organized by the MIPT Digital Department.
Coached students for the Artificial Intelligence track of the National Technology Olympiad.
Mentored participants of the National Technology Olympiad, supporting them in solving AI and data-related challenges.
Mentored a team in the NLP Pro educational program by the Fintech Hub of the Bank of Russia and VTB.
Led a team in an educational program focused on no-code approaches to text clustering.
Worked as an NLP engineer in an early-stage startup, focusing on text analytics and user feedback analysis.
Joined a Skoltech research group after completing the SMILES Machine Learning Summer School, continuing the research project initiated during the school.
Worked on a machine learning project focused on time-series clustering.
Reviewed participants’ code submissions and provided feedback during the AI-ARROW bootcamp.
A collection of practical notebooks created as part of my educational and outreach activities. The project covers various useful techniques, from accelerating neural network training to exploring interesting approaches in data analysis and machine learning workflows.
An ongoing educational project with structured text lectures and notes based on my learning and teaching experience. The project is designed to make data science and machine learning topics more accessible and easier to revisit.
A course project developed for Deep Learning School, focused on helping students build simple MVP-style applications using Streamlit. The materials guide learners through turning machine learning ideas into interactive prototypes.
Educational materials for a seminar on clustering prepared within the MIPT Deep Learning School. The project includes practical examples and explanations of clustering methods used in machine learning and data analysis.
A research-oriented project focused on improving Retrieval-Augmented Generation systems by incorporating additional information about embedding quality. The project explores how retrieval quality signals can be used to make RAG pipelines more reliable and effective.
M. M. Kotyga. Published in the proceedings of the 67th All-Russian Scientific Conference at MIPT, “Applied Mathematics and Informatics”, Moscow, March 31 — April 5, 2025, pp. 248–249.
M. M. Kotyga. Accepted to the proceedings of “Information Technology, Data Science and Artificial Intelligence: From Theory to Applied Solutions”, Tashkent, December 11–12, 2025. In press.
M. M. Kotyga. Published in the proceedings of the XVII All-Russian Scientific and Technical Conference with International Participation “Robotics and Artificial Intelligence”, Zheleznogorsk, December 13, 2025, pp. 209–213.
Menschikov M. et al. A research preprint studying model-specific and language-specific position effects in multilingual large language models.
M. M. Kotyga. Accepted to the proceedings of the 68th All-Russian Scientific Conference at MIPT, “Applied Mathematics and Informatics”, Moscow, March 30 — April 4, 2026. In press.
An interview about my experience as a mentor at the National Technology Olympiad, supporting participants and helping them work through technology-focused challenges.
Participation in the final stage of the FINOPOLIS.365 Youth Program hackathon, presenting a solution developed as part of the competition.
Presentation of the results of the educational program by the Fintech Hub of the Bank of Russia and VTB at the Data Fusion conference.
An introductory session for the Middle Track of the AI-ARROW Bootcamp, focused on the general approach to solving machine learning competition tasks.
A workshop session dedicated to discussing the solution strategy for a machine learning competition task, including practical considerations and modeling decisions.
A seminar on Retrieval-Augmented Generation, covering the core ideas behind RAG systems, their architecture, and practical aspects of building retrieval-based LLM applications.
Presented research on methods for improving the efficiency of Retrieval-Augmented Generation systems for information retrieval by incorporating embedding quality assessment into the retrieval pipeline.