Maxim Kryzhanovskiy

Designed and developed an AutoML tool for building classification models based on the Transformer Architecture. Developed MLOps pipelines using TFX, Airflow, and Kubernetes. Worked on the alignment of LLMs with PEFT methods (LoRA, DoRA) using instruct-tuning, RLHF, and DPO.

Researched and developed methods for summarization of scientific articles. Built an AI-agent system for autonomous scientific reviews based on user queries using LangChain, smolagents, and vLLM. Researched and developed methods in scientific text and image generations with diffusion models.

Managed the development of AI-powered solutions for business process optimization. Designed and managed the development of a Computer Vision system integrated into a robotic arm for autonomous construction work. Designed and managed the development of an MES-system for food factories, integrated into over 20 factories.

Developed a sound stack (pulse and alsa) for a secure operating system using C/C++. Integrated sound stack backends into different devices using a special Building Kit. Worked with CMake and JTest for CI/CD.

Activities and societies: Academic advisor: Prof. Konstantin Vorontsov. Research area: Efficient ways of tuning Large Language Models in high domains with big contexts and high requirements for precision.

Studied applied mathematics and informatics with a focus on computational methods. Researched efficient ways of tuning Large Language Models in high domains with big contexts and high requirements for precision.