AI Agents · Multilingual Systems · RAG

Building reliable agents for long-horizon interaction.

I work on AI agent systems that combine tool use, multilingual dialogue, domain reasoning, and memory-augmented retrieval. My current interests include reliable long-horizon agents, multimodal RAG, and post-training optimization.

View Projects View CV Read Notes

Selected Work

Three representative directions that show both engineering execution and research potential.

Multilingual Customer Service Agent

Built agent workflows for overseas business scenarios, covering user intent, domain knowledge, tool calling, and multilingual response quality.

Agent Tool Use Evaluation

Reasoning Data Distillation

Constructed reasoning-focused datasets through model distillation, rule-based cleaning, domain validation, and failure analysis.

CoT SFT DPO

Memory-augmented Multimodal RAG

Research direction on persistent memory, multimodal retrieval, and reliable task completion across long interactions.

RAG Memory MLLM

What this site is for

A clean evidence layer behind my CV: selected projects, technical writing, research interests, and contact links.

For recruiters

Quickly understand my role, project scope, technical stack, and execution ability.

For professors

See my research direction, industrial context, and potential fit for AI agent research.

For collaborators

Find my notes, interests, and open questions around reliable AI systems.