Key Contributions:- Developed and integrated an LLM solution (Llama2-70b, LangChain, RAG) for a major international appliance manufacturer, analyzing over 1 million customer support chats to extract topics, solutions, and customer satisfaction levels, reducing support response time by 25%.- Conducted research on advanced prompting techniques, improving LLM reasoning and information retrieval accuracy in RAG systems by 30%, leading to more relevant and accurate outputs.- Developed and deployed an ML-based prediction system for one of Eastern Europe's largest leasing companies processing data from 5 million clients, optimizing data pipelines and increasing new client prediction accuracy by 20%, which contributed to a significant boost in sales conversions.

Key contributions:- Developed business chatbots — for medicine, HR, media, NGOs — using LLMs, RAG, LangChain and NLP techniques; deployed as microservices- Fine-tuned LLaMA 7B-13B, Mistral 7B, Zephyr; optimized LLM inference on GPU- Developed CV-based software (PyTorch) for efficient multi-face detection (RetinaFace) & anonymization in videos, optimized for mobile GPUs

Key Contributions:- Developed a data-driven web app for transforming real-time sensory data into digital art (Python, FastAPI, Express.js, Docker)

Key contributions:- Designed and implemented an end-to-end system for emotion recognition from facial expressions (CNN) and brain activity (SVM, LSTM), achieving 80-95% accuracy- Developed a microservice architecture using FastAPI, Docker, Kafka; created data pipelines for EEG and facial images; deployed models on AWS- Built interactive dashboards (Matplotlib) to visualize emotional dynamics; set up a CI/CD pipeline (GitHub Actions) to streamline model updates

Worked in the Institute of Cognitive Neuroscience of the Higher School of Economics,Ranked №1 university in Russia, colloquially referred to as "Russian MIT".Key contributions: - Developed an end-to-end pipeline for automated pose extraction from video and movement features synchronization using OpenPose, OpenCV, NumPy, SciPy, Pandas- Built pipelines and classifiers (LDA, SVM, CNN, GAN models) for EEG data in Brain-Computer Interfaces with scikit-learn and PyTorch- Used RL, Q-learning for modelling neural networks in the brain- Built efficient data pipelines for collection, preprocessing, analysis, and visualization leveraging Python, R, and MATLAB- Led the design and execution of scientific experiments. Analyzed the effectiveness of various ML architectures and computational techniques in cognitive neuroscience- Composed and contributed to scientific papers to document findings and advancements- Worked extensively with health-related data, including EEG, EMG, Eye-tracking, GSR, and fMRI

Worked in the Laboratory for Experimental Urban DesignKey contributions:- Leveraged Natural Language Processing techniques to analyze social media posts, identifying usage patterns, visitor attitudes, and public perception of various urban areas- Constructed Python-based interfaces for the collection and visualization of geospatial data. Handled extensive anonymized datasets, preparing them for advanced ML analyses- Managed research laboratories, orchestrated cross-disciplinary research, and led the design of educational programs and public events focused on cutting-edge data science for urban studies