Barlas Oğuz
Research Scientist at Meta
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolBarlas Oğuz is a research scientist in the San Francisco Bay Area with 11 years of experience building production-scale AI systems across industry and academia. At Meta he leads modeling efforts for LLM chatbots and generative 3D models for the Metaverse, and has built neural retrieval, QA, and confidence estimation systems that serve billions of requests daily. Earlier work at Microsoft included speech and language modeling for Cortana and contributions to CNTK, and his academic background (Ph.D. EECS, UC Berkeley) underpins work on long-range dependent stochastic models and stable distributed protocols. An early co-creator and contributor to the open-source PyText NLP framework, he combines deep theoretical rigor with hands-on engineering to optimize GPU memory, multitask model exports, and loss-weighting strategies in production pipelines.
11 years of coding experience
9 years of employment as a software developer
Bachelor of Science (BS), Electrical and Electronics Engineering, Bachelor of Science (BS), Electrical and Electronics Engineering at Bilkent UniversityDoctor of Philosophy (Ph.D.), Electrical Engineering and Computer Sciences, Doctor of Philosophy (Ph.D.), Electrical Engineering and Computer Sciences at UC Berkeley
Turkish, English, German, Spanish
Github Skills (12)
model-building10
pytorch10
nlp10
backend10
python10
modeling10
model-driven10
model-driven-development10
optimization9
gpu9
optimisation9
machine-learning9
Programming languages (3)
Github contributions (5)
facebookresearch/pytext
Dec 2018 - Dec 2019
A natural language modeling framework based on PyTorch
Role in this project:
Backend Developer
Backend DeveloperContributions:56 commits, 59 PRs, 12 comments in 11 months
Contributions summary:Barlas primarily contributed to the `pytext` repository, a natural language modeling framework based on PyTorch. Their work focused on implementing features related to model saving, batch processing, and model export, including enabling individual model exports in a disjoint multitask setting. Key contributions also involved optimizing GPU memory usage during training and implementing loss weighting across different tasks within the framework.
pytorchnlpbertmachine-learningnatural-language
borguz/pytext-1
Dec 2018 - Nov 2019
A natural language modeling framework based on PyTorch
Contributions:148 pushes, 57 branches in 11 months
pytorchnlpbertmachine-learningnatural-language
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.