Nikita Durasov
Research Scientist, VLM Post-Training & Agents at NVIDIA
Wollerau, Schwyz, Switzerland
Summary
👤
Senior🎓
Top SchoolNikita Durasov is a research scientist with a decade of experience building production-grade AI that transforms massive unstructured video and image data into searchable, actionable knowledge using vision-language models, multimodal LLMs, RAG, and knowledge-graph driven search. Currently at NVIDIA, he leads VLM fine-tuning, video data mining and labeling pipelines, and has a track record of taking research ideas from concept through deployment at scale. His background includes PhD research at EPFL on uncertainty, Bayesian methods and active learning, plus applied roles at Apple, Amazon Prime Air, Samsung and top Russian research institutes where he improved detectors, depth estimation and large-scale dataset operations. Fluent across the stack from data collection and active learning to model post-training and systems integration, he combines academic rigor with engineering pragmatism. A less obvious strength is his consistent focus on data-efficiency—using Bayesian and active learning techniques to reduce labeling needs while improving model robustness.
10 years of coding experience
7 years of employment as a software developer
Doctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at EPFLBachelor's degree (with honours) Applied Math and Physics (Yandex chair), Bachelor's degree (with honours) Applied Math and Physics (Yandex chair) at Moscow Institute of Physics and Technology (State University) (MIPT)Master’s-level program in Computer Science and Data Analysis Machine Learning Developer, Master’s-level program in Computer Science and Data Analysis Machine Learning Developer at Yandex School of Data Analysis
Русский, Английский, Французский, Немецкий
Stackoverflow
11reputation
125reached
1answer
0questions
Github Skills (26)
uncertainty-quantification9
paper9
robustness9
uncertainty9
computer-vision8
corruption8
generalization8
datasets7
tensorflow7
deep-learning7
machine-learning6
pytorch6
google-cloud-platform6
bayesian-inference5
idempotent5
Programming languages (3)
Github contributions (5)
nikitadurasov/torch-ttt
Jan 2025 - Nov 2025
A modular and easy-to-use framework for Test-Time Training (TTT) and Test-Time Adaptation (TTA) in Pytorch, making your networks more generalizable with minimal effort ✨
Contributions:8 PRs, 64 pushes, 4 branches in 10 months
pytorchgeneralizationrobustnesstest-time-adaptationtest-time-training
nikitadurasov/masksembles
Mar 2021 - Jun 2021
Official repository for the paper "Masksembles for Uncertainty Estimation" (CVPR 2021).
Contributions:28 commits, 3 PRs, 19 pushes in 3 months
pytorchcomputer-visionestimationuncertaintyuncertainty-estimation
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.