Fields of interest: Multimodal, Generative AI (Bayesian, Diffusion, VAE/GAN, Optimal Transport, Flow Matching), Embodied AI (Planning, Manipulation, Imitation learning)

My telegram channel: AI - Arseny Ivanov

My curriculum vitae: CV

• Sberbank, Sber Robotics Lab / NLP Multimodal Researcher (Feb 2024 - Present)
  • Developed a FeedBack Module for evaluating of the execution of actions in the environment (Video-Text Retrieval - SFT CLIP, ViClip, VideoLlava, VideoMamba).
• Sberbank, Sber AmazMe / Research Data Scientist / Recommender Systems (Jun 2022 - Feb 2024):
  • Active participation in deploying algorithms into production operation
  • Conducting and analyzing A/B tests
  • Reading papers from RecSys conferences, participating in the internal reading club
  • Managed to significantly improve the recommendations: preprocessing and filtering of data, optimization of model training
• Sberbank, Sber AI / Research Assistant (Jan 2022 - Nov 2022)
  • Conducted an analysis of advanced scientific papers on the problem of predicting travel time
  • Developed graph algorithms for travel time estimation, improved predictions quality by 11%
  • Published research paper TTE on GNN for the JBD ”Journal of Big Data”
• Sberbank, Sber Devices / Data Science Intern (Oct 2021 - Jan 2022)
  • Research in the field of neuromorphic artificial intelligence and cognitive systems of the brain
  • Developed a binary classifier of Covid19 diseases in humans based on data from an electromagnetic cap with sensors

• Skolkovo Institute of Science And Technology "Skoltech" and Higher School of Economics "HSE" (2024 - Present)
  • MCs of two diplomas, Data Science, Mathematics of Machine Learning (MML)
• National University of Science and Technology "MISIS" (2020 - 2024)
  • BCs, Institute of Computer Science, Applied Mathematics
  • Diploma thesis: Action Planning by Multimodal-LLM models

• Support program of commercially oriented researches ”UMNIK” (NUST MISIS / Sber Robotics Lab)
  • Learning LLM models for the tasks of robotic manipulation 2022-2023 | Video | Poster | (6500$ award)

• Gct-TTE: graph convolutional transformer for travel time estimation. Journal of Big Data, 11(1), 15. 2023

• 🥇 Digital Breakthrough "Video Recommendation System for RuTube" 2023 | Solution | Diploma | (5000$ award)
• 🥇 Agora Hackathon "Marketplace service powered by AI" 2022 | Solution | Diploma | (2500$ award)
• 🥇 Rosatom City Hackathon "Business Analytics on Graphs alghoritms" 2021 | Diploma | (2000$ award)
• 🥇 «Breakthrough Energy» Hackathon "Predictive system for gas distribution system" 2021 | Diploma | (3000$ award)
• 🥇 Whoosh Mobility Hackathon "TTE and Route Search service using GNN" 2021 | Solution | Diploma | (2500$ award)
• 🥈 Avanpost Hackathon "Transfer learning for ResNet50" 2022 | Solution | Diploma | (1000$ award)
• 🥈 Leaders of Digital Transformation «LDT» "TTE and Route Prediction" 2023 | Solution | Diploma | (6000$ award)
• 4/40 X-MAS Hackathon "Named-entity recognition (NER) with documents" 2022 | Solution
• 5/30 Sber Zvuk Hackathon "CV-deblurring popular personalities on videoo" 2021 | Diploma
• 7/45 Lenta Hack The Cart "Recommendation System for marketplace Lenta" 2021 | Diploma
• 5/30 AgroCode Hackathon "Building an optimized driving route" 2021 | Diploma
• 78/501 MTS ML Cup | Solution | Diploma
• 1/4 final ICPC: The 2020 Moscow Regional Contest | Diploma

• AIRI Summer School ”Summer with AIRI” (St. Petersburg, Russia, 2024)
  • Poster: Vision-Language Models for Robot Success Detection
• International Conference & Summer School ”ISSCAI” by HSE (St. Petersburg, Russia, 2024)
  • Poster: Graph Convolutional Transformers for Travel Time Estimation
• AIRI Summer School ”Summer with AIRI” (Kazan, Russia, 2023)
  • Poster: Grounding Local LLMs to Robotic Manipulation Tasks
  • Habr paper: How to set a list of actions using language models and voice commands?
• Head of DS Community AI Knowledge Club. Lecturer of the courses:
  • Random forest, Ensembles, Bagging, Boosting - Video
  • Introduction to Deep Learning (Tensors, MLP, CNN) - Video
  • Unsupervised ML (KMeans, DBSCAN, SVD, PCA, tSNE, UMAP) - Video
  • Transformers and Attention (Self-Attention, Encoder, Decoder) - Video

• Binary VQA: binary answer to the question on the image (BLIP, VILT, RoBERTa + VIT)
• NLP Projects: implementing Transformer, BERT analyzing, NER with BERT
• LLM Roles: KL Divergence between Multivariate Gaussians of LLM roles

• Astra Linux Scholarship holder (2023 - 2024)
• VK Education Scholarship holder (2023 - 2024)

🇷🇺 Russian - Native
🇬🇧 English - С1 (Intermediate-Advanced)
🇨🇵 French - A2 (Elementary)