An LLM researcher who gazes at the starlit skies of Artificial General Intelligence ✨
Ex‑intern at Microsoft Research Asia (MSRA) Systems & Networking Group mentored by Dr. Li Lyna Zhang.
Previously research assistant at UIUC with Prof. Fan Lai and Prof. Minjia Zhang.
Incoming M.S. in Intelligent Information Systems @ CMU LTI in Fall 2025.
Open to MLE and SDE intern opportunities!
📂 (~/about_me/)
Hi there! 👋 I’m Gaokai Zhang. I am an MIIS at CMU‑LTI since Fall 2025 (NLP/LLM) and I hold dual B.S. degrees at ZJUI (CompE @ UIUC, ECE @ ZJU).
From July 2024 to July 2025, I joined Microsoft SRG group as an intern. At MSRA, I gained hands‑on experience with LLMs, RL, SFT, and business‑scale ML systems. I incorporated novel data synthesis and solid reinforcement learning in (LoongRL) to make 7B models surpass 32B LRMs in long-context reasoning tasks at even 100k-200k tokens. I contributed LongRoPE2—extending LLM context windows to 128 K tokens while retaining 98.5 % short‑context accuracy (ICML 2025 poster)—and built a parallel pipeline for large‑scale user‑query processing and delivered production‑ready long‑context recommendation models to Microsoft Asia‑Pacific R&D.
At UIUC, I have worked on:
- Monte‑Carlo‑Tree‑Search planning for cost‑efficient LLM training on heterogeneous GPUs/TPUs.
- Robustness benchmarking of LLMs (Stochastic Monkeys).
🧠 what_drives_me.txt
- Long‑context reasoning & scaling
- Cloud‑optimized training architectures
- Fine-tuning and reinforcement learning
Outside work I’ve logged 8 000 + h gaming 🎮 (also a lifetime Faker fan), vibe to rap 🎧, and hunt the perfect omakase bite 🍣.
Feel free to reach out, always happy to connect with like‑minded friends and collaborators!
🏅 highlights.json
- 🎓 MIIS @ CMU LTI (2025 – 2027)
- 🧪 ICML 2025 poster: LongRoPE2 (arXiv)
- 💼 Research Intern @ MSRA
- 🏫 Dual B.S. CompE @ UIUC & ECE @ ZJU
📰 changelog.md
- 🧾 May 2025 – LongRoPE2 accepted at ICML 🎉
- 🏢 Jul 2024 – Joined MSRA as intern
- 🧑🔬 Mar 2024 – Started LLM projects @ UIUC