Education

University of Bern
Ph. D in Computer Science, advised by Prof. Dr. Paolo Favaro.
Research Topics: Learning Object Interactions, Controllable Video Generation, World Models.
Sep 2020 - present
Lomonosov Moscow State University
B.S. + M.S. equivalent (with honors) in Mathematics, advised by Prof. Dr. Igor Sheipak.
Thesis: Spectral analysis of the operator in the generalized ’t Hooft model.
Sep 2014 - June 2020
Yandex School of Data Analysis
M.S. equivalent in Data Analysis, advised by Prof. Dr. Ilya Muchnik.
Thesis: Optimization of oil production via forecast models on injection wells operation modes.
Sep 2014 - June 2020

Employment

Yandex
Software Engineer at the Backend of Yandex.Navigator. Implemented algorithms for (sub)optimal route proposals on road graphs. Improved the quiality of the service via machine learning.
Feb 2019 - Sep 2020
Moscow, Russia
Inline Group
ML Specialist. Developed oil production forecast models.
Sep 2017 - Dec 2021
Moscow, Russia

Publications

Enabling Visual Composition and Animation in Unsupervised Video Generation
A. Davtyan, S. Sameni, B. Ommer, P. Favaro
A model to compose and animate scenes from sparse sets of visual features.
arXiv, 2024
Learn the Force We Can: Enabling Sparse Motion Control in Multi-Object Video Generation
A. Davtyan, P. Favaro
A model to animate single frames with sparse motion control.
AAAI, 2024
Multi-View Unsupervised Image Generation with Cross Attention Guidance Single image to novel view synthesis without any supervision.
arXiv, 2023
Efficient Video Prediction via Sparsely Conditioned Flow Matching
A. Davtyan, S. Sameni, P. Favaro
Conditioning only on a few randomly chosen past frames at each denoising step of flow matching results into a more efficient training procedure.
ICCV, 2023
Controllable Video Generation through Global and Local Motion Dynamics
A. Davtyan, P. Favaro
A model to discover agents' action spaces from a dataset of videos in an unsupervised way. The action spaces are decomposed into global (2D shifts) and local (discrete) actions.
ECCV, 2022
KOALA: A Kalman Optimization Algorithm with Loss Adaptivity A neural network optimization algorithm based on Kalman filtering.
AAAI, 2022