Sotiris Anagnostidis
Member Of Technical Staff at Anthropic
Switzerland
Summary
🤩
Rockstar🎓
Top SchoolSotiris Anagnostidis is a Member of Technical Staff at Anthropic with nine years of experience bridging cutting-edge ML research and production engineering. He holds a PhD in Data Science from ETH Zurich and strong academic foundations from NTUA, combining rigorous research with hands-on model training and evaluation work. His contributions to high-profile open-source projects like LAION-AI/Open-Assistant include reward model and RL training pipeline improvements, dialogue data collators, and refactors that sharpen training loss computation. Prior roles at Meta, Google DeepMind, and ETH Zurich show a pattern of short, impactful research engagements and collaboration across industry and academia. Known for meticulous implementation details and a knack for improving reproducibility in ML pipelines, he brings both theoretical depth and practical delivery to large-scale conversational AI systems.
9 years of coding experience
1 year of employment as a software developer
Arsakeio Psychikou9.90, Electrical and Computer Engineering, 9.90/10.00, 9.90, Electrical and Computer Engineering, 9.90/10.00 at National Technical University of AthensMaster's degree, Data Science, 5.86/6.00, Master's degree, Data Science, 5.86/6.00 at ETH Zurich
Github Skills (12)
transformers10
pytorch10
machine-learning10
trainings10
python10
modeling10
evaluation9
eval9
loss-functions8
ai8
openai7
language-model7
Programming languages (2)
Github contributions (5)
LAION-AI/Open-Assistant
Jan 2023 - Mar 2023
OpenAssistant is a chat-based assistant that understands tasks, can interact with third-party systems, and retrieve information dynamically to do so.
Role in this project:
ML Engineer
ML EngineerContributions:89 reviews, 50 commits, 59 PRs in 2 months
Contributions summary:Sotiris contributed to the model training and evaluation pipelines. Their work included implementing features for reward model training, such as integrating output directories, and made corrections to SFT model loading. They also worked on RL training by implementing prompt selection and configuration, as well as refining the data collator for dialogue-based datasets. Additionally, they refactored code related to loss calculations.
language-modelpythonpartyassistantchat
sanagno/sanagno.github.io
Jan 2022 - Dec 2024
Contributions:8 pushes, 1 branch in 2 years 11 months
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.