Research I develop algorithmic (and typically ML-based) solutions for the most important and intriguing problems. At Herophilus (biotech startup growing "brains-in-a-dish" called organoids) built data pipelines and analysed TL imaging, single-cell sequencing, IF organoid slices, ICC images, video imaging of neural activity and gene expression profiles. Developed models/approaches for vision in company, introduced an efficient and reliable algorithm to learn genotypes during demultiplexing in single-cell sequencing. At Intelligence Lab (Samsung Research) investigated various approaches to one-shot learning for face recognition on mobile devices. Led development of neural text-to-speech synthesis, developed a data-efficient soft alignment of speech and text. Also worked on object detection, active learning, and model conversion for production At Yandex/CERN developed ML approaches to different problems in high energy physics at the Large Hadron Collider. Developed machine learning-based approaches for particle identification, tracking, online data filtering, flavour tagging and particle shower detection. My methods were primarily developed for the LHCb experiment, but later spread across other CERN experiments. Previous research topics include optimal control, mathematical physics, and even solid state theory. Deep technical background helps in design of efficient algorithms and approaches, while wide experience allows dealing with non-trivial aspects of algorithms' existence-in-the-wild. And critical thinking is crucial to build proper validation.