Focused on building production-ready post-training pipelines — from data curation and SFT to RLHF and alignment — with hands-on experience across three AI organizations.
|
AI Research Intern · Kingsoft 2026.03 — Present SFT · RLHF · Alignment Iterating post-training pipelines for production LLMs — data quality control, preference-pair construction, and alignment behavior optimization. |
|
AI Research Intern · Zhipu AI 2025.09 — 2026.02 SFT · RL · Evaluation Built SFT + RL pipelines covering instruction formatting, training recipe tuning, and post-training evaluation with inference-oriented validation. |
|
AI Research Intern (Multimodal) · China Telecom AI Research Institute 2025.06 — 2025.09 Multimodal · Fine-tuning · Evaluation Multimodal R&D across dataset preparation, fine-tuning, and evaluation for text-image understanding; bridged research prototypes with serving pipelines. |
