每一条主张都能回到同一个事实库

第一页：ML projects + evaluation + engineering

• Lead with a measurable model result and a clear baseline.证据：portfolio-timeseries
• Show reproducible validation and honest limitations.证据：experiment-design
• Make the implementation stack visible on page one.证据：engineering-stack

第一页：LLM + RAG + evaluation + deployment

• Lead with retrieval evaluation rather than model-name lists.证据：retrieval-evaluation
• Use the product case to explain trade-offs and feedback loops.证据：career-os-product

第一页：Statistics + time-series + reproducible research

• Lead with the time-series experiment and its baseline.证据：portfolio-timeseries
• Emphasize leakage prevention and out-of-sample discipline.证据：experiment-design
• Make Python and SQL immediately visible.证据：engineering-stack

第一页：AI literacy + impact + leadership + communication

• Turn a technical experiment into a decision and impact narrative.证据：experiment-design
• Present Career Decision OS as a product case study.证据：career-os-product
• Show scope, trade-offs, and stakeholder alignment.证据：cross-functional-delivery