SpaceX Super IntelligenceMember of Technical Staff
Dec 2025 — PresentSingapore
Working on coding models and agents, with a focus on data-centric training and evaluation.
Qian Liu coding @xai, previously tiktok and sea ai lab. opinions are my own
Qian Liu coding @xai, previously tiktok and sea ai lab. opinions are my own
Career
TikTokCoding Team Lead
Nov 2024 — Dec 2025Singapore
Built a 10+ person coding research team. Focus on code pre-training and agentic reinforcement learning.
Sea AI LabResearch Scientist
Jul 2022 — Nov 2024Singapore
Led the development of the Sailor LLM family, with the full pre-training pipeline: data crawling and cleaning, deduplication, data mixture optimization (RegMix), and continual pre-training.
Worked with
Cosigns Received
Featured Media
x.ai
Introducing Grok 4.6
Grok 4.6 builds on Grok 4.5 with a particular focus on long-running agents and more ambitious interactive and visual work.
x.ai
Introducing Grok 4.5
Grok 4.5 is SpaceXSI's smartest model built for coding, agentic tasks, and knowledge work.
x.com
Scaling Laws with Vocabulary: Larger Models Deserve Larger Vocabularies
We investigate how vocabulary size impacts language model scaling laws by training models ranging from 33M to 3B parameters on up to over 500B characters with various vocabulary configurations. We find that the optimal vocabulary depends on the compute budget, thus we propose three approaches to determine it.
opencoder-llm.github.io
OpenCoder: Top-Tier Open Code Large Language Models
OpenCoder is an open and reproducible code LLM family which matches the performance of Top-Tier Code LLM. We provide not just the final models, but also the reproducible training data, the complete data processing pipeline, rigorous experimental ablation results, and detailed training protocols for open scientific research.
