Denis leads scientific and engineering research in computer vision, multimodal learning, and generative artificial intelligence. At Sber AI, he led the team that trained the first Russian text2image transformation models ruDALL-E 1.3B (Malevich) and ruDALL-E 12B (Kandinsky 1.0), diffusion models Kandinsky 2.0, 2.1, 2.2, bimodal multitasking model RUDOLPH, GHOST for one-shot deepfake generation, and Digital Peter for handwriting recognition. Denis actively publishes research results in journals and abstracts of international conferences.