Volodymyr Kuleshov is a Cornell computer-science professor and co-founder and chief scientist of Inception. He studies generative models, uncertainty and applications in science, health and sustainability, including diffusion-based language models.[1][2]
Start with Applied Machine Learning →
Education
Links
Website
X 
GitHub Applied Machine Learning notes Joan Eliasoph, M.D. Assistant Professor, Computer Science; Cornell Tech · Cornell University
Co-founder and chief scientist · Inception
Career
Kuleshov studied mathematics and computer science at McGill and earned his doctorate at Stanford. Cornell announced his faculty appointment in July 2018. His university website describes earlier work on genome sequencing commercialized through Moleculo and co-founding Afresh, which applies AI to reducing food waste. His current account identifies his scientific leadership role at Inception.[3][1][2]
Contributions
Block Diffusion, coauthored by Kuleshov and collaborators, combines sequential generation with diffusion inside blocks of text. Block size gives a way to move between autoregressive and diffusion approaches. The paper develops training and sampling methods, including caching, to address limitations in sequence length and inference. Its experiments compare quality and efficiency under specified conditions rather than establishing that every diffusion model is faster.[4]
Tools, writing & talks
Writing, research & talks
Start here
Applied Machine Learningkuleshov-group.github.ioOpen course notes on implementing, evaluating and improving machine-learning systems.
Start with the introductory chapters, then implement a baseline for a small dataset. Basic Python and some probability help; use the exercises to connect methods with evaluation.
- Block Diffusionarxiv.org
A method that interpolates between autoregressive generation and diffusion within text blocks.
Co-author[4]