AI Engineering · Multimodal Learning · Agent Workflows

I build AI applications and multimodal pipelines, with a focus on reproducible experiments, reliable workflows, and maintainable systems.

专注 AI 应用与多模态系统,关注从实验验证到工程实现的完整过程。

• Multimodal Learning — Cross-modal representation learning, alignment, and retrieval.
• LLM Applications — Agent workflows, RAG, tool calling, and evaluation.
• AI Engineering — Modular Python systems, configuration-driven workflows, and reproducible environments.

A training and evaluation framework for multimodal person re-identification across five modalities.

• Combines CLIP with modality-aware LoRA routing and semantic alignment.
• Supports multimodal sampling, training, evaluation, and submission generation.
• Emphasizes training stability and consistency between training and evaluation.

• Translate ambiguous requirements into clear tasks and system designs.
• Use evaluation and feedback to guide iteration.
• Prioritize reproducibility, reliability, and maintainability.

Core tools: Python · Linux · Git

Build systems that can be tested, understood, and improved.

LingmaFuture's GitHub statistics

Most used languages

AI / ML

• LLM Applications · Agentic Workflows · Multimodal Systems
• RAG · Tool Calling · Evaluation & Feedback Loops

Engineering

• Python-centric systems
• Modular design · Config-driven architecture
• Containerized & reproducible workflows

_{Build systems, not demos.
Optimize for leverage, not noise.}