hello there! ✨

Ekaterina Antipushina

Екатерина Антипушина

NeuroAI · Embodied AI · LLMs · Computer Vision

I work on NeuroAI, Embodied AI, large language models (LLMs), and computer vision. I develop multimodal models for brain signals and images, and LLM agents that understand 3D scenes, use tools, and reason about spatial relationships.

PhD student at Skoltech · ML Engineer at Applied AI Institute

Ekaterina Antipushina — NeuroAI and Embodied AI researcher
Google Scholar ORCID Skoltech profile ResearchGate

About Me

Ekaterina Antipushina is a researcher and ML engineer working on NeuroAI, Embodied AI, large language models (LLMs), and computer vision.She is a PhD student at Skoltech and a Machine Learning Engineer at the Spatial Intelligence Lab, Applied AI Institute. Her work connects multimodal representation learning and generative modeling with applications to neural signals, medical images, and 3D scenes.

In NeuroAI, her work covers brain activity modeling, EEG and fMRI integration, cross-modal prediction, and real-time neurofeedback. Her projects include pyOpenNFT, an open-source framework for neurofeedback with machine learning, and CSTNet, a generative approach to EEG-to-ECoG mapping using optimal transport.

In Embodied AI and computer vision, she develops LLM agents and vision-language systems for 3D scene understanding, language-guided object localization, and spatial reasoning. Her engineering work includes retrieval-augmented generation (RAG), tool calling, multi-step answer verification, and automated evaluation. Details are available in her CV.

She also develops LLM applications for research, including Neuroforum, a platform with MCP tools, background agent workflows, and model usage accounting. Her broader research includes medical computer vision, interpretable multimodal biomedical analysis, brain organoids, and biomarkers. See her publications on Google Scholar.

Outside the lab, you'll find me in Moscow's cozy cafés seeking inspiration.

My Research Interests

NeuroAI

  • Neural representation learning and foundation models
  • EEG, fMRI, and cross-modal prediction
  • Generative modeling and optimal transport
  • Real-time neurofeedback

Embodied AI & Spatial Intelligence

  • Vision-language models for 3D scenes
  • Language-guided object localization
  • Scene graphs and spatial relationships
  • Evaluation of spatial reasoning

Multimodal & Biomedical AI

  • Spatiotemporal modeling and data fusion
  • Medical image analysis
  • Interpretable machine learning
  • Biomedical signals and biomarker discovery

Professional Journey

explore my background...

Academic CV · PDF · September 2026

My Journey

the path that led me here...

Machine Learning Engineer
Spatial Intelligence Lab, Applied AI Institute
October 2025 – present
Developing LLM agents and computer vision systems for 3D scene understanding in a joint project with Huawei. Combining vision-language models, scene graphs and geometry to locate objects from natural-language queries. Building retrieval-augmented (RAG) and tool-calling workflows, multi-step answer verification, and automated evaluation of visual grounding and spatial reasoning.
PhD student
Skolkovo Institute of Science and Technology (Skoltech)
2024 – present
Research on multimodal neural data and real-time prediction of localized brain activity. Developed tokenization and representation-learning methods for EEG and fMRI, including masked pretraining, Transformers, and discrete representations.
Research Engineer, Medical AI
Skoltech · Neuroimaging and Cognitive Neuroscience Lab
November 2024 – September 2025
Developed pyOpenNFT, an open-source Python framework for real-time fMRI and EEG-fMRI neurofeedback. Implemented parallel signal processing with shared memory and a FastAPI prediction service for integrating machine-learning models. First author of the MICCAI 2025 framework paper.
Machine Learning Engineer · part-time
BIMAI-Lab, University of Sharjah
July 2024 – August 2025
Led computational analysis of transcriptomic and proteomic data in international studies of human brain organoids under hypoxia. Developed statistical and machine-learning workflows, interpreted results with the experimental team, and contributed to papers in Fluids and Barriers of the CNS and iScience.
ML Engineer / Research Engineer
Skoltech · Medical AI / BIMAI-Lab
January 2023 – December 2024
Developed generative and interpretable models for multimodal neuroimaging and biomedical data, including Rest2Task, EEG-to-ECoG modeling, schizophrenia classification, and kidney-cancer metastasis prediction. Worked with data fusion, feature selection, graph representations, and validation on small medical datasets.
Master of Science
Skoltech
2022 – 2024
Research focus: medical AI and multimodal neuroimaging. Thesis on interpretable machine-learning models and data fusion for the analysis of neuroimaging and biomedical data in schizophrenia.
Bachelor’s degree in Biomedical Engineering
Moscow Aviation Institute
2018 – 2022
Engineering foundations for biomedical technology, physiological measurements, and medical systems.

Research Hub

explore my research journey...

Projects Papers Posters Talks

Projects

3D Spatial Understanding
Embodied AI · Spatial Intelligence
Developing LLM agents and computer vision systems for 3D scene understanding in a joint project with Huawei. Combining vision-language models, scene graphs and geometry to locate objects from natural-language queries. Building retrieval-augmented (RAG) and tool-calling workflows, multi-step answer verification, and automated evaluation of visual grounding and spatial reasoning.
Vision-Language Models 3D Grounding Scene Graphs

Project details in my CV

Neural Representation Learning
NeuroAI · EEG & fMRI
Developed tokenization and pretraining methods for EEG and fMRI using Transformer/ViT architectures, BSQ and VQ-VAE representations, and masked modeling. Evaluated transfer across subjects and recording configurations with controls for data leakage.
Transformers Masked Modeling EEG / fMRI
Medical Image Modeling
2026 · MRI, Ultrasound & OCT
Research on cardiac MRI segmentation and biomarker estimation, transfer-aware ultrasound biometry, and inverse OCT phantom generation. Three accepted contributions to MICCAI 2026 workshops and challenges.
Segmentation Transfer Learning Inverse Rendering

Read the papers

Brain Organoid Data Analysis
Biomedical AI · Transcriptomics & Proteomics
Computational analysis of molecular responses to hypoxia in human brain and choroid plexus organoids. Statistical modeling and integration of omics data supported studies published in Fluids and Barriers of the CNS (2025) and iScience (2026).
Statistical Modeling Omics Biomarkers

Read the papers

pyOpenNFT
Real-time neurofeedback · MICCAI 2025
Open-source Python framework for real-time fMRI and EEG-fMRI neurofeedback with machine-learning integration, parallel signal processing, and a prediction service.
Python FastAPI Neurofeedback

Code on GitHub

CSTNet
EMA4MICCAI 2025 · proceedings 2026
Generative EEG-to-ECoG mapping using optimal transport. Contributed to the framework described in the EMA4MICCAI 2025 paper, published by Springer in 2026.
PyTorch Optimal Transport EEG-to-ECoG

Code on GitHub

Rest2Task
Generative neuroimaging
Generative framework for predicting task-based fMRI from resting-state recordings using variational autoencoders and generative adversarial models.
VAE GAN fMRI

Code on GitHub

Neuroforum
Research software · AI agents
Designed a platform for researchers and AI agents working with scientific materials. Implemented a FastAPI service, MCP tools, access controls, background jobs with RabbitMQ/Dramatiq, LLM cost accounting, Docker Compose, and CI checks.
FastAPI MCP Docker

Publications

Journal and conference papers, accepted work, and preprints.

Google Scholar CV · PDF

2026

2025

Posters & Presentations

Rest2Task: generative modeling for task-based fMRI prediction

Sber RnD Day · 2024

View poster

The application of machine learning methods for multimodal neuroimaging data in the diagnosis of schizophrenia

LIFT Conference · 2024

View poster

Development of a personalized Transcranial Magnetic Stimulation complex utilizing biofeedback

LIFT School of Young Neurotechnologists · 2024

View poster
Research illustration · original poster unavailable

Benchmarking of Machine Learning and Deep Learning approaches for Neuroimaging Data

Summer of Machine Learning · Skoltech · 2023

Publication record

Invited talks & presentations

Invited talk

Universal Brain Encoder: The Evolution of Models for Neural Signals

NeuroTalk Meetup · MISIS University, Moscow

Event
About the talk

An invited talk at the NeuroTalk meetup on neurotechnology and brain-computer interfaces. The talk traces how models for neural signals evolved from task-specific architectures towards universal brain encoders — foundation models trained across subjects, montages and recording modalities — and what this shift means for decoding and for brain-computer interfaces.

Conference talk

Bridge Between Time and Space: How Generative AI Turns EEG into fMRI

DataFest · Open Data Science

Recording Slides
About the talk

The report is devoted to the application of generative models for the transformation of neuroimaging data with high temporal resolution (EEG) into functional and spatial representations (fMRI), combining their strengths. Modern algorithms, architectural solutions and examples of use in diagnostics and brain research are considered

Let's Connect!

I love connecting with fellow researchers, collaborators, and anyone passionate about making AI solutions!

coffee chats welcome! ☕

Whether you want to discuss research ideas, explore collaboration opportunities, or just share thoughts about the future of AI in healthcare, I'd love to hear from you.

Email Me

ekantipushina@gmail.com

best way to reach me! 📧

LinkedIn

katherine-antipushina

let's connect professionally! 💼

GitHub

utoprey

code & collaborations! 💻

📍 Based in Moscow, Russia

ML Engineer at Applied AI Institute · PhD student at Skoltech

always up for meeting fellow researchers in Moscow! 🇷🇺