Eugen Hotaj
Member Of Technical Staff at Perplexity
Redwood City, California, United States
Summary
🤩
Rockstar🎓
Top SchoolEugen Hotaj is a Member of Technical Staff with 12 years of experience building and optimizing machine learning systems across Google, Meta, Perplexity, and Microsoft AI, focused on AutoML, neural architecture search, LLMs, and agentic reinforcement learning. He combines deep mathematical rigor from an NYU Mathematics and Computer Science degree with a practical engineering mindset—often digging into low-level estimator and TPU compatibility issues to squeeze performance from models in production. Notable open-source work includes refactoring TensorFlow AdaNet’s TPUEstimator to improve TPU support and summaries, reflecting a talent for making research-grade ML frameworks production-ready. Based in Redwood City, he gravitates toward post-training and agentic RL problems, bridging mid- and post-training model stages to enable scalable, search-augmented and autonomous model behaviors. Colleagues describe him as someone who prefers to truly understand systems end-to-end rather than accept opaque abstractions.
12 years of coding experience
9 years of employment as a software developer
Bachelor’s Degree Mathematics and Computer Science, Bachelor’s Degree Mathematics and Computer Science at New York University
Albanian
Stackoverflow
383reputation
10kreached
24answers
0questions
Github Skills (15)
machine-learning10
deeplearning-ai10
deep-learning10
tensorflow10
python10
tpu10
distributed-training9
arraylist6
pandas6
javascript6
bytecode6
jvm6
scala6
binary-search-tree6
java6
Programming languages (7)
Github contributions (5)
tensorflow/adanet
Nov 2018 - Nov 2019
Fast and flexible AutoML with learning guarantees.
Role in this project:
ML Engineer
ML EngineerContributions:71 commits, 1 push, 10 comments in 1 year
Contributions summary:Eugen primarily contributed to the AdaNet project by refactoring and modifying the TPUEstimator component. These changes included adapting the estimator for TPU compatibility, improving summary handling, and enhancing the model's performance when using TPU. The user's contributions focused on enabling and optimizing the use of TPUs within the AdaNet framework.
automlmachine-learningflexibledeep-learningdistributed-training
EugenHotaj/talko
Apr 2020 - Mar 2021
A client-server chat app in Python.
Contributions:54 commits, 3 PRs, 49 pushes in 11 months
pythonchat-appclient-serverchatchat-server
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.