AI applications · Data engineering · Data analytics
Xiangguo Zhang
I build reliable data pipelines, decision tools, and AI applications that show how they work and where they stop.
Open to Data Analytics, Data Engineering, and AI Application Engineering roles.
06case studies03disciplines6 interactive demos
AI applicationsI build AI applications with repeatable checks, human approval, and clear operating limits.Data engineeringI break pipelines on purpose, then check recovery against source, table, and event state.Data analyticsI build decision tools from governed data and make every assumption visible.
Selected work
Open any project to see what I built, how it works, and how to verify it.
01Release GuardianAI application / release engineeringA 13-node LangGraph release gate: four evidence retrievers, deterministic validators, bounded retries — and the publish decision stays with a human.132 live graph runs8/8 aggregate gates passed100% citation fidelityLive evaluation + demo02RAG Quality LabDeterministic RAG evaluationA controlled regression test caught a knowledge-base update degrading the strongest pipeline; I then extended the same evidence lifecycle into tracked evaluation infrastructure for 11,309 enterprise documents.4/12 questions regressed11,309 docs130 enterprise questionsData and test baseline03Privacy Preflight WebBrowser-local redaction workbenchA browser-local workbench for reviewing sensitive text, images, and multi-page PDFs with English + Simplified Chinese OCR and destructive export checks.96 embedded-worker tests67 recorded browser casesOCR 19/19 hits / 2 false positivesLocal browser workflow04Margin Control TowerAnalytics engineering / margin decisionsWhen weekly contribution margin changes, this browser tool starts from a hash-verified Olist aggregate, traces discounts, returns, cost of goods, and fulfillment, then tests a bounded promotion scenario before recording an action.15,809 Olist aggregate rows99,441 source orders10 fail-closed contract checksInteractive decision workbench05Streaming Reliability LabMySQL CDC → Flink → Iceberg · failure injectionI break a MySQL-to-Flink-to-Iceberg pipeline on purpose, then check that source state, table snapshots, and event IDs still agree after recovery.5 induced failure classes0 snapshot diffs after recoveryRecorded recovery replay06Credit Policy LabRisk analytics / policy governanceThis browser lab starts from a hash-verified Lending Club backtest and keeps offline scores separate from expected-loss math, policy thresholds, review capacity, monitoring, and the final recorded decision.120,000 scored loans24,000 later backtest rowscapacity-gated policy auditScore-to-policy simulator