Researcher in machine learning systems, large language models, and distributed systems.
How media typically covers Hao Zhang
Directly quoted in these articles
Google introduces Tunix, a JAX-native open-source library that simplifies LLM post-training workflows including fine-tuning, alignment, and distillation, achieving ~12% relative improvement on math reasoning benchmarks.
“Quoted praising Tunix as an ideal lightweight framework for post-training reinforcement learning with gaming environments”