Applied AI engineer building retrieval-grounded assistants, agent workflows, and practical LLM products.

I work on systems that connect language models to real knowledge and real actions: hybrid retrieval, semantic indexing, citations, tool calling, stateful workflows, evaluation, and production integration.

An AI reading assistant for the Model Within knowledge site. It supports article summaries, selection-based Q&A, and site-wide retrieval. The public repository is a documentation-first case study; the production implementation remains private.

A Streamlit application that turns a topic, target duration, and creativity setting into a structured Chinese video script. It supports multiple OpenAI-compatible model providers and augments generation with Chinese Wikipedia retrieval.

• RAG systems: hybrid retrieval, reranking, semantic search, citations, and retrieval evaluation
• Agent systems: tool calling, stateful orchestration, failure recovery, and observable execution
• Applied LLM products: multi-provider integration, privacy boundaries, and user-facing workflows
• Knowledge engineering: turning experiments into reproducible notes, demos, and production features

I prefer systems that are measurable, debuggable, and honest about their boundaries. A model response is useful only when the data path, failure path, and verification path are clear.

Writing and experiments: modelwithin.cloud