I work across AI research, high-performance computing, and large-scale data infrastructure, turning research ideas into systems that actually run.
My work spans large-scale dataset construction, efficient inference, high-performance computing, and machine learning. I am particularly interested in the intersection of computation, intelligence, and scalable systems.
Built tooling for a large-scale dataset of 2ch, a major Russian imageboard. The project collects publicly accessible threads, metadata, and multimedia content to construct a machine-readable snapshot of the Russian Internet.
View projectA continuously evolving collection of optimized LLM inference and decoding techniques, tested on European supercomputers including Leonardo. The project explores techniques such as KVPress, prompt caching, graph inference, and prompt lookup, with implementations for both conventional environments and HPC clusters.
View projectReimplemented an existing Kolmogorov complexity based image classification approach as a modern Python library. The project provides a compression-based alternative to conventional neural approaches, particularly relevant to classification problems where training data is limited.
View project