Zhilin Wang

Senior Applied Scientist at NVIDIA LLM Alignment

Email at wangzhilin123 [at] gmail [dot] com

Hi there! I’m Zhilin and I work on Aligning LLMs to be more helpful for humans! My recent work focuses on leading the HelpSteer dataset series (1, 2 and 3), which are popular open-source post-training datasets for helpfulness modelling. Then, I explore creative ways to model such data, giving rise to the Llama-3.1-Nemotron-70B-Instruct and Reward models, as well as the Llama-3.3-Nemotron-70B model family.

I also do some hobby projects in my personal capability outside of work - such as Humanoid Agents (Platform for simulation for human-like agents).

Highlights

Oct 2024 Built HelpSteer2-Preference & Llama-3.1-Nemotron-70B models w/ 1M+ downloads (ICLR 2025)
Aug 2023 Architected Humanoid Agents Simulation Platform (EMNLP 2023 System Demo) GitHub stars
Jan 2017 Started an EdTech startup helping 500k+ Chemistry students on learnAh.org
Aug 2015 Top 20 Global Finalist at Google Science Fair with my project on Zinc-Air Batteries šŸ”‹
Jan 2014 Sliver Medalist at Singapore Chemistry Olympiad (SChO) and invited for IChO training team 🄈