Artem Chumachenko
AI Engineer at Together AI
Amsterdam, North Holland, Netherlands
Summary
🤩
Rockstar🎓
Top SchoolArtem Chumachenko is an AI engineer with a decade of experience building and fine-tuning large language models and dialog systems, currently contributing at Together AI from Amsterdam. He has driven production and research efforts at Yandex and Neiro.ai—training YaLM variants, improving generative-retrieval personalization, and launching chatbot features—grounded in an applied physics and math background from MIPT. Artem combines research instincts (inventing compression techniques like "Weight Squeezing") with practical systems work, including implementing generation mixins, beam search, and adapter support for the popular petals distributed LLM platform. He focuses on speeding up fine-tuning and inference for large transformers and bringing experimental models into real-world conversational products.
10 years of coding experience
4 years of employment as a software developer
Bachelor's degree, Applied Physics and Mathematics, Bachelor's degree, Applied Physics and Mathematics at Moscow Institute of Physics and Technology (State University) (MIPT)
Github Skills (11)
transformer10
machine-learning10
pytorch10
nlp10
language-model10
python10
large-language-models9
distributed-systems9
gpt9
deep-learning9
mixture7
Programming languages (4)
Github contributions (5)
bigscience-workshop/petals
Jun 2022 - Jan 2023
🌸 Run LLMs at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading
Role in this project:
ML Engineer
ML EngineerContributions:30 reviews, 97 commits, 40 PRs in 6 months
Contributions summary:Artem primarily contributed to the development of a distributed language model platform, specifically focusing on integrating generation capabilities within the `petals` framework. Their contributions include implementing a `RemoteGenerationMixin` class for auto-regressive text generation, adding support for various decoding algorithms such as greedy search and sampling, and integrating prefix-tuned inference. The user also introduced the `beam_search` algorithm and designed functionality for utilizing adapters. Their work directly impacts the models capabilities for generation tasks.
llama2finetuningbittorrentllamahuggingface-transformers
catalyst-team/dl-course
Sep 2020 - Dec 2020
Deep Learning with Catalyst
Contributions:68 commits, 47 PRs, 47 pushes in 3 months
deep-learningcatalystmachine-learning
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.